fix roi index, clean up pose, roi, records, vector

9 files changed, 524 insertions, 100 deletions

diff --git a/megapixels/commands/cv/face_pose.py b/megapixels/commands/cv/face_pose.py
index c37d006f..9979ad34 100644
--- a/megapixels/commands/cv/face_pose.py
+++ b/megapixels/commands/cv/face_pose.py
@@ -1,5 +1,9 @@
 """
 Converts ROIs to pose: yaw, roll, pitch
+pitch: looking down or up in yes gesture
+roll: tilting head towards shoulder
+yaw: twisting head left to right in no gesture
+
 """
 
 import click
@@ -17,7 +21,7 @@ from app.settings import app_cfg as cfg
   help='Override enum media directory')
 @click.option('--data_store', 'opt_data_store',
   type=cfg.DataStoreVar,
-  default=click_utils.get_default(types.DataStore.SSD),
+  default=click_utils.get_default(types.DataStore.HDD),
   show_default=True,
   help=click_utils.show_help(types.Dataset))
 @click.option('--dataset', 'opt_dataset',
@@ -91,50 +95,57 @@ def cli(ctx, opt_fp_in, opt_fp_out, opt_dir_media, opt_data_store, opt_dataset,
   # store poses and convert to DataFrame
   poses = []
 
-  # iterate 
+  # iterate groups with file/record index as key
   for record_index, df_img_group in tqdm(df_img_groups):
     # make fp
     ds_record = df_record.iloc[record_index]
     fp_im = data_store.face(ds_record.subdir, ds_record.fn, ds_record.ext)
     im = cv.imread(fp_im)
-    # get bbox
-    x = df_img_group.x.values[0]
-    y = df_img_group.y.values[0]
-    w = df_img_group.w.values[0]
-    h = df_img_group.h.values[0]
-    dim = im.shape[:2][::-1]
-    bbox = BBox.from_xywh(x, y, w, h).to_dim(dim)
-    # get pose
-    landmarks = face_landmarks.landmarks(im, bbox)
-    pose_data = face_pose.pose(landmarks, dim, project_points=opt_display)
-    pose_degrees = pose_data['degrees']  # only keep the degrees data
-
-    # use the project point data if display flag set
-    if opt_display:
-      pts_im = pose_data['points_image']
-      pts_model = pose_data['points_model']
-      pt_nose = pose_data['point_nose']
-      dst = im.copy()
-      face_pose.draw_pose(dst, pts_im, pts_model, pt_nose)
-      face_pose.draw_degrees(dst, pose_degrees)
-      # display to cv window
-      cv.imshow('', dst)
-      while True:
-        k = cv.waitKey(1) & 0xFF
-        if k == 27 or k == ord('q'):  # ESC
-          cv.destroyAllWindows()
-          sys.exit()
-        elif k != 255:
-          # any key to continue
-          break
-
-    # add image index and append to result CSV data
-    pose_degrees['record_index'] = record_index
-    poses.append(pose_degrees)
+    # iterate image group dataframe with roi index as key
+    for roi_index, df_img in df_img_group.iterrows():
+      # get bbox
+      x, y, w, h = df_img.x, df_img.y, df_img.w, df_img.h
+      dim = (ds_record.width, ds_record.height)
+      #dim = im.shape[:2][::-1]
+      bbox = BBox.from_xywh(x, y, w, h).to_dim(dim)
+      # get pose
+      landmarks = face_landmarks.landmarks(im, bbox)
+      pose_data = face_pose.pose(landmarks, dim)
+      #pose_degrees = pose_data['degrees']  # only keep the degrees data
+      #pose_degrees['points_nose'] = pose_data
+      # use the project point data if display flag set
+      if opt_display:
+        dst = im.copy()
+        face_pose.draw_pose(dst, pose_data['point_nose'], pose_data['points'])
+        face_pose.draw_degrees(dst, pose_data)
+        # display to cv window
+        cv.imshow('', dst)
+        while True:
+          k = cv.waitKey(1) & 0xFF
+          if k == 27 or k == ord('q'):  # ESC
+            cv.destroyAllWindows()
+            sys.exit()
+          elif k != 255:
+            # any key to continue
+            break
 
+      # add image index and append to result CSV data
+      pose_data['roi_index'] = roi_index
+      for k, v in pose_data['points'].items():
+        pose_data[f'point_{k}_x'] = v[0][0] / dim[0]
+        pose_data[f'point_{k}_y'] = v[0][1] / dim[1]
+      pose_data.pop('points')
+      pose_data['point_nose_x'] = pose_data['point_nose'][0] / dim[0]
+      pose_data['point_nose_y'] = pose_data['point_nose'][1] / dim[1]
+      pose_data.pop('point_nose')
+      poses.append(pose_data)
 
-  # save date
+  # create dataframe
   file_utils.mkdirs(fp_out)
   df = pd.DataFrame.from_dict(poses)
+  # save date
   df.index.name = 'index'
-  df.to_csv(fp_out)
\ No newline at end of file
+  df.to_csv(fp_out)
+  # save script
+  cmd_line = ' '.join(sys.argv)
+  file_utils.write_text(cmd_line, '{}.sh'.format(fp_out))
\ No newline at end of file
diff --git a/megapixels/commands/cv/face_pose_mt.py b/megapixels/commands/cv/face_pose_mt.py
new file mode 100644
index 00000000..8fef2c2c
--- /dev/null
+++ b/megapixels/commands/cv/face_pose_mt.py
@@ -0,0 +1,138 @@
+"""
+Converts ROIs to pose: yaw, roll, pitch
+"""
+
+import click
+
+from app.settings import types
+from app.utils import click_utils
+from app.settings import app_cfg as cfg
+
+@click.command()
+@click.option('-i', '--input', 'opt_fp_in', default=None,
+  help='Override enum input filename CSV')
+@click.option('-o', '--output', 'opt_fp_out', default=None,
+  help='Override enum output filename CSV')
+@click.option('-m', '--media', 'opt_dir_media', default=None,
+  help='Override enum media directory')
+@click.option('--data_store', 'opt_data_store',
+  type=cfg.DataStoreVar,
+  default=click_utils.get_default(types.DataStore.SSD),
+  show_default=True,
+  help=click_utils.show_help(types.Dataset))
+@click.option('--dataset', 'opt_dataset',
+  type=cfg.DatasetVar,
+  required=True,
+  show_default=True,
+  help=click_utils.show_help(types.Dataset))
+@click.option('--size', 'opt_size', 
+  type=(int, int), default=(300, 300),
+  help='Output image size')
+@click.option('--slice', 'opt_slice', type=(int, int), default=(None, None),
+  help='Slice list of files')
+@click.option('-f', '--force', 'opt_force', is_flag=True,
+  help='Force overwrite file')
+@click.option('-d', '--display', 'opt_display', is_flag=True,
+  help='Display image for debugging')
+@click.pass_context
+def cli(ctx, opt_fp_in, opt_fp_out, opt_dir_media, opt_data_store, opt_dataset, opt_size, 
+  opt_slice, opt_force, opt_display):
+  """Converts ROIs to pose: roll, yaw, pitch"""
+  
+  import sys
+  import os
+  from os.path import join
+  from pathlib import Path
+  from glob import glob
+  
+  from tqdm import tqdm
+  import numpy as np
+  import dlib  # must keep a local reference for dlib
+  import cv2 as cv
+  import pandas as pd
+
+  from app.models.bbox import BBox
+  from app.utils import logger_utils, file_utils, im_utils
+  from app.processors.face_landmarks import LandmarksDLIB
+  from app.processors.face_pose import FacePoseDLIB
+  from app.models.data_store import DataStore
+
+  # -------------------------------------------------
+  # init here
+
+  log = logger_utils.Logger.getLogger()
+  
+  # set data_store
+  data_store = DataStore(opt_data_store, opt_dataset)
+
+  # get filepath out
+  fp_out = data_store.metadata(types.Metadata.FACE_POSE) if opt_fp_out is None else opt_fp_out
+  if not opt_force and Path(fp_out).exists():
+    log.error('File exists. Use "-f / --force" to overwite')
+    return
+
+  # init face processors
+  face_pose = FacePoseDLIB()
+  face_landmarks = LandmarksDLIB()
+
+  # load filepath data
+  fp_record = data_store.metadata(types.Metadata.FILE_RECORD)
+  df_record = pd.read_csv(fp_record).set_index('index')
+  # load ROI data
+  fp_roi = data_store.metadata(types.Metadata.FACE_ROI)
+  df_roi = pd.read_csv(fp_roi).set_index('index')
+  # slice if you want
+  if opt_slice:
+    df_roi = df_roi[opt_slice[0]:opt_slice[1]]
+  # group by image index (speedup if multiple faces per image)
+  df_img_groups = df_roi.groupby('record_index')
+  log.debug('processing {:,} groups'.format(len(df_img_groups)))
+
+  # store poses and convert to DataFrame
+  poses = []
+
+  # iterate 
+  for record_index, df_img_group in tqdm(df_img_groups):
+    # make fp
+    ds_record = df_record.iloc[record_index]
+    fp_im = data_store.face(ds_record.subdir, ds_record.fn, ds_record.ext)
+    im = cv.imread(fp_im)
+    for roi_id, df_img in df_img_group.iterrows():
+      # get bbox
+      x, y, w, h = df_img.x, df_img.y, df_img.w, df_img.h
+      dim = im.shape[:2][::-1]
+      bbox = BBox.from_xywh(x, y, w, h).to_dim(dim)
+      # get pose
+      landmarks = face_landmarks.landmarks(im, bbox)
+      pose_data = face_pose.pose(landmarks, dim, project_points=opt_display)
+      pose_degrees = pose_data['degrees']  # only keep the degrees data
+
+      # use the project point data if display flag set
+      if opt_display:
+        pts_im = pose_data['points_image']
+        pts_model = pose_data['points_model']
+        pt_nose = pose_data['point_nose']
+        dst = im.copy()
+        face_pose.draw_pose(dst, pts_im, pts_model, pt_nose)
+        face_pose.draw_degrees(dst, pose_degrees)
+        # display to cv window
+        cv.imshow('', dst)
+        while True:
+          k = cv.waitKey(1) & 0xFF
+          if k == 27 or k == ord('q'):  # ESC
+            cv.destroyAllWindows()
+            sys.exit()
+          elif k != 255:
+            # any key to continue
+            break
+
+      # add image index and append to result CSV data
+      pose_degrees['record_index'] = record_index
+      poses.append(pose_degrees)
+
+
+  # save date
+  file_utils.mkdirs(fp_out)
+  df = pd.DataFrame.from_dict(poses)
+  df.index.name = 'index'
+  df.to_csv(fp_out)
\ No newline at end of file
diff --git a/megapixels/commands/cv/face_roi.py b/megapixels/commands/cv/face_roi.py
index a08566a8..a09a1ce0 100644
--- a/megapixels/commands/cv/face_roi.py
+++ b/megapixels/commands/cv/face_roi.py
@@ -20,7 +20,7 @@ color_filters = {'color': 1, 'gray': 2, 'all': 3}
   help='Override enum media directory')
 @click.option('--data_store', 'opt_data_store',
   type=cfg.DataStoreVar,
-  default=click_utils.get_default(types.DataStore.SSD),
+  default=click_utils.get_default(types.DataStore.HDD),
   show_default=True,
   help=click_utils.show_help(types.Dataset))
 @click.option('--dataset', 'opt_dataset',
@@ -52,10 +52,12 @@ color_filters = {'color': 1, 'gray': 2, 'all': 3}
   help='Filter to keep color or grayscale images (color = keep color')
 @click.option('--largest/--all-faces', 'opt_largest', is_flag=True, default=True,
   help='Only keep largest face')
+@click.option('--zone', 'opt_zone', default=(0.0, 0.0), type=(float, float), 
+  help='Face center must be located within zone region (0.5 = half width/height)')
 @click.pass_context
 def cli(ctx, opt_fp_in, opt_dir_media, opt_fp_out, opt_data_store, opt_dataset, opt_size, opt_detector_type, 
   opt_gpu, opt_conf_thresh, opt_pyramids, opt_slice, opt_display, opt_force, opt_color_filter,
-  opt_largest):
+  opt_largest, opt_zone):
   """Converts frames with faces to CSV of ROIs"""
   
   import sys
@@ -130,7 +132,7 @@ def cli(ctx, opt_fp_in, opt_dir_media, opt_fp_out, opt_data_store, opt_dataset,
         continue
         
     try:
-      bboxes = detector.detect(im, size=opt_size, pyramids=opt_pyramids, largest=opt_largest)
+      bboxes = detector.detect(im, size=opt_size, pyramids=opt_pyramids, largest=opt_largest, zone=opt_zone)
     except Exception as e:
       log.error('could not detect: {}'.format(fp_im))
       log.error('{}'.format(e))
@@ -142,17 +144,17 @@ def cli(ctx, opt_fp_in, opt_dir_media, opt_fp_out, opt_data_store, opt_dataset,
         'x': bbox.x, 
         'y': bbox.y, 
         'w': bbox.w, 
-        'h': bbox.h,
-        'image_width': im.shape[1],
-        'image_height': im.shape[0]}
+        'h': bbox.h
+        }
       data.append(roi)
+    if len(bboxes) == 0:
+      log.warn(f'no faces in: {fp_im}')
     
     # debug display
     if opt_display and len(bboxes):
       im_md = im_utils.resize(im, width=min(1200, opt_size[0]))
       for bbox in bboxes:
         bbox_dim = bbox.to_dim(im_md.shape[:2][::-1])
-        log.debug(f'bbox: {bbox_dim}')
         cv.rectangle(im_md, bbox_dim.pt_tl, bbox_dim.pt_br, (0,255,0), 3)
       cv.imshow('', im_md)
       while True:
@@ -168,4 +170,7 @@ def cli(ctx, opt_fp_in, opt_dir_media, opt_fp_out, opt_data_store, opt_dataset,
   file_utils.mkdirs(fp_out)
   df = pd.DataFrame.from_dict(data)
   df.index.name = 'index'
-  df.to_csv(fp_out)
\ No newline at end of file
+  df.to_csv(fp_out)
+  # save script
+  cmd_line = ' '.join(sys.argv)
+  file_utils.write_text(cmd_line, '{}.sh'.format(fp_out))
\ No newline at end of file
diff --git a/megapixels/commands/cv/face_vector.py b/megapixels/commands/cv/face_vector.py
index 7200d73b..7c03205c 100644
--- a/megapixels/commands/cv/face_vector.py
+++ b/megapixels/commands/cv/face_vector.py
@@ -15,7 +15,7 @@ from app.settings import app_cfg as cfg
   help='Override enum media directory')
 @click.option('--data_store', 'opt_data_store',
   type=cfg.DataStoreVar,
-  default=click_utils.get_default(types.DataStore.SSD),
+  default=click_utils.get_default(types.DataStore.HDD),
   show_default=True,
   help=click_utils.show_help(types.Dataset))
 @click.option('--dataset', 'opt_dataset',
@@ -90,36 +90,28 @@ def cli(ctx, opt_fp_out, opt_dir_media, opt_data_store, opt_dataset, opt_size,
   log.debug('processing {:,} groups'.format(len(df_img_groups)))
 
   vecs = []
-
-  for image_index, df_img_group in tqdm(df_img_groups):
+  for record_index, df_img_group in tqdm(df_img_groups):
     # make fp
-    roi_index = df_img_group.index.values[0]
-    # log.debug(f'roi_index: {roi_index}, image_index: {image_index}')
-    ds_file = df_record.loc[roi_index]  # locate image meta
-    #ds_file = df_record.loc['index', image_index]  # locate image meta
-
-    fp_im = data_store.face(str(ds_file.subdir), str(ds_file.fn), str(ds_file.ext))
+    ds_record = df_record.iloc[record_index]
+    fp_im = data_store.face(ds_record.subdir, ds_record.fn, ds_record.ext)
     im = cv.imread(fp_im)
-    # get bbox
-    x = df_img_group.x.values[0]
-    y = df_img_group.y.values[0]
-    w = df_img_group.w.values[0]
-    h = df_img_group.h.values[0]
-    imw = df_img_group.image_width.values[0]
-    imh = df_img_group.image_height.values[0]
-    dim = im.shape[:2][::-1]
-    # get face vector
-    dim = (imw, imh)
-    bbox_dim = BBox.from_xywh(x, y, w, h).to_dim(dim)  # convert to int real dimensions
-    # compute vec
-    # padding=opt_padding not yet implemented in 19.16 but merged in master
-    vec = facerec.vec(im, bbox_dim, jitters=opt_jitters)
-    vec_str = ','.join([repr(x) for x in vec])  # convert to string for CSV
-    vecs.append( {'roi_index': roi_index, 'record_index': image_index, 'vec': vec_str})
+    for roi_index, df_img in df_img_group.iterrows():
+      # get bbox
+      x, y, w, h = df_img.x, df_img.y, df_img.w, df_img.h
+      dim = (ds_record.width, ds_record.height)
+      #dim = im.shape[:2][::-1]
+      # get face vector
+      bbox_dim = BBox.from_xywh(x, y, w, h).to_dim(dim)  # convert to int real dimensions
+      # compute vec
+      # padding=opt_padding not yet implemented in 19.16 but merged in master
+      vec = facerec.vec(im, bbox_dim, jitters=opt_jitters)
+      vec_str = ','.join([repr(x) for x in vec])  # convert to string for CSV
+      vecs.append( {'roi_index': roi_index, 'record_index': record_index, 'vec': vec_str})
 
 
-  # save date
+  # create dataframe
   df = pd.DataFrame.from_dict(vecs)
   df.index.name = 'index'
+  # save CSV
   file_utils.mkdirs(fp_out)
   df.to_csv(fp_out)
\ No newline at end of file
diff --git a/megapixels/commands/cv/face_vector_mt.py b/megapixels/commands/cv/face_vector_mt.py
new file mode 100644
index 00000000..412f9806
--- /dev/null
+++ b/megapixels/commands/cv/face_vector_mt.py
@@ -0,0 +1,118 @@
+"""
+Converts ROIs to face vector
+"""
+
+import click
+
+from app.settings import types
+from app.utils import click_utils
+from app.settings import app_cfg as cfg
+
+@click.command()
+@click.option('-o', '--output', 'opt_fp_out', default=None,
+  help='Override enum output filename CSV')
+@click.option('-m', '--media', 'opt_dir_media', default=None,
+  help='Override enum media directory')
+@click.option('--data_store', 'opt_data_store',
+  type=cfg.DataStoreVar,
+  default=click_utils.get_default(types.DataStore.SSD),
+  show_default=True,
+  help=click_utils.show_help(types.Dataset))
+@click.option('--dataset', 'opt_dataset',
+  type=cfg.DatasetVar,
+  required=True,
+  show_default=True,
+  help=click_utils.show_help(types.Dataset))
+@click.option('--size', 'opt_size', 
+  type=(int, int), default=(300, 300),
+  help='Output image size')
+@click.option('-j', '--jitters', 'opt_jitters', default=cfg.DLIB_FACEREC_JITTERS,
+  help='Number of jitters')
+@click.option('-p', '--padding', 'opt_padding', default=cfg.DLIB_FACEREC_PADDING,
+  help='Percentage padding')
+@click.option('--slice', 'opt_slice', type=(int, int), default=(None, None),
+  help='Slice list of files')
+@click.option('-f', '--force', 'opt_force', is_flag=True,
+  help='Force overwrite file')
+@click.option('-g', '--gpu', 'opt_gpu', default=0,
+  help='GPU index')
+@click.pass_context
+def cli(ctx, opt_fp_out, opt_dir_media, opt_data_store, opt_dataset, opt_size, 
+  opt_slice, opt_force, opt_gpu, opt_jitters, opt_padding):
+  """Converts face ROIs to vectors"""
+  
+  import sys
+  import os
+  from os.path import join
+  from pathlib import Path
+  from glob import glob
+  
+  from tqdm import tqdm
+  import numpy as np
+  import dlib  # must keep a local reference for dlib
+  import cv2 as cv
+  import pandas as pd
+
+  from app.models.bbox import BBox
+  from app.models.data_store import DataStore
+  from app.utils import logger_utils, file_utils, im_utils
+  from app.processors import face_recognition
+
+
+  # -------------------------------------------------
+  # init here
+
+  log = logger_utils.Logger.getLogger()
+  # set data_store
+  data_store = DataStore(opt_data_store, opt_dataset)
+
+  # get filepath out
+  fp_out = data_store.metadata(types.Metadata.FACE_VECTOR) if opt_fp_out is None else opt_fp_out
+  if not opt_force and Path(fp_out).exists():
+    log.error('File exists. Use "-f / --force" to overwite')
+    return
+
+  # init face processors
+  facerec = face_recognition.RecognitionDLIB()
+
+  # load data
+  fp_record = data_store.metadata(types.Metadata.FILE_RECORD)
+  df_record = pd.read_csv(fp_record).set_index('index')
+  fp_roi = data_store.metadata(types.Metadata.FACE_ROI)
+  df_roi = pd.read_csv(fp_roi).set_index('index')
+
+  if opt_slice:
+    df_roi = df_roi[opt_slice[0]:opt_slice[1]]
+  
+  # -------------------------------------------------
+  # process here
+  df_img_groups = df_roi.groupby('record_index')
+  log.debug('processing {:,} groups'.format(len(df_img_groups)))
+
+  vecs = []
+  for record_index, df_img_group in tqdm(df_img_groups):
+    # make fp
+    ds_record = df_record.iloc[record_index]
+    fp_im = data_store.face(ds_record.subdir, ds_record.fn, ds_record.ext)
+    im = cv.imread(fp_im)
+    for roi_index, df_img in df_img_group.iterrows():
+      # get bbox
+      x, y, w, h = df_img.x, df_img.y, df_img.w, df_img.h
+      imw = df_img.image_width
+      imh = df_img.image_height
+      dim = im.shape[:2][::-1]
+      # get face vector
+      dim = (imw, imh)
+      bbox_dim = BBox.from_xywh(x, y, w, h).to_dim(dim)  # convert to int real dimensions
+      # compute vec
+      # padding=opt_padding not yet implemented in 19.16 but merged in master
+      vec = facerec.vec(im, bbox_dim, jitters=opt_jitters)
+      vec_str = ','.join([repr(x) for x in vec])  # convert to string for CSV
+      vecs.append( {'roi_index': roi_index, 'record_index': record_index, 'vec': vec_str})
+
+
+  # save date
+  df = pd.DataFrame.from_dict(vecs)
+  df.index.name = 'index'
+  file_utils.mkdirs(fp_out)
+  df.to_csv(fp_out)
\ No newline at end of file
diff --git a/megapixels/commands/datasets/records.py b/megapixels/commands/datasets/file_record.py
index b6ef618b..355b22f2 100644
--- a/megapixels/commands/datasets/records.py
+++ b/megapixels/commands/datasets/file_record.py
@@ -10,7 +10,12 @@ from app.utils.logger_utils import Logger
 
 log = Logger.getLogger()
 
-identity_sources = ['subdir', 'subdir_head', 'subdir_tail']
+# Choose part of the filepath that will be used for the person identity
+# eg subdirectory "lfw/media/original/batch_1/train/barack_obama/001.jpg" --> [subdir_tail] --> "barack_obama"
+# eg subdirectory "lfw/media/original/batch_1/train/barack_obama/001.jpg" --> [subdir_head] --> "batch_1"
+# eg subdirectory "lfw/media/original/batch_1/train/barack_obama/001.jpg" --> [subdir] --> "barack_obama"
+
+identity_sources = ['subdir', 'numeric']
 
 @click.command()
 @click.option('-i', '--input', 'opt_fp_in', default=None,
@@ -21,7 +26,7 @@ identity_sources = ['subdir', 'subdir_head', 'subdir_tail']
   help='Override enum media directory')
 @click.option('--data_store', 'opt_data_store',
   type=cfg.DataStoreVar,
-  default=click_utils.get_default(types.DataStore.SSD),
+  default=click_utils.get_default(types.DataStore.HDD),
   show_default=True,
   help=click_utils.show_help(types.Dataset))
 @click.option('--dataset', 'opt_dataset',
@@ -35,7 +40,8 @@ identity_sources = ['subdir', 'subdir_head', 'subdir_tail']
   help='Number of threads')
 @click.option('-f', '--force', 'opt_force', is_flag=True,
   help='Force overwrite file')
-@click.option('--identity', 'opt_identity', default=None, type=click.Choice(identity_sources),
+@click.option('--identity', 'opt_identity', type=click.Choice(identity_sources),
+  default='numeric',
   help='Identity source, blank for no identity')
 @click.option('--recursive/--no-recursive', 'opt_recursive', is_flag=True, default=False,
   help='Use glob recursion (slower)')
@@ -44,7 +50,7 @@ def cli(ctx, opt_fp_in, opt_fp_out, opt_dataset, opt_data_store, opt_dir_media,
   opt_identity, opt_force, opt_recursive):
   """Generates sha256, uuid, and identity index CSV file"""
   
-  import sys
+  import sys, os
   from glob import glob
   from os.path import join
   from pathlib import Path
@@ -53,9 +59,11 @@ def cli(ctx, opt_fp_in, opt_fp_out, opt_dataset, opt_data_store, opt_dir_media,
   import random
   import uuid
 
+  import cv2 as cv
   import pandas as pd
   from tqdm import tqdm
   from glob import glob
+  from operator import itemgetter
 
   from app.models.data_store import DataStore
   from app.utils import file_utils, im_utils
@@ -91,15 +99,26 @@ def cli(ctx, opt_fp_in, opt_fp_out, opt_dataset, opt_data_store, opt_dir_media,
 
   pbar = tqdm(total=len(fp_ims))
 
-  def as_sha256(fp_im):
+  def pool_mapper(fp_im):
     pbar.update(1)
-    return file_utils.sha256(fp_im)
+    sha256 = file_utils.sha256(fp_im)
+    im = cv.imread(fp_im)
+    w, h = im.shape[:2][::-1]
+    file_size_kb = os.stat(fp_im).st_size // 1000
+    num_channels = im_utils.num_channels(im)
+    return {
+      'width': w, 
+      'height': h, 
+      'sha256': sha256,
+      'file_size_kb': file_size_kb, 
+      'num_channels': num_channels
+      }
 
   # convert to thread pool
-  sha256s = []  # ?
+  pool_maps = []  # ?
   pool = ThreadPool(opt_threads) 
   with tqdm(total=len(fp_ims)) as pbar:
-    sha256s = pool.map(as_sha256, fp_ims)
+    pool_maps = pool.map(pool_mapper, fp_ims)
   pbar.close()
 
 
@@ -108,11 +127,12 @@ def cli(ctx, opt_fp_in, opt_fp_out, opt_dataset, opt_data_store, opt_dir_media,
   
   data = []
   indentity_count = 0
-  for sha256, fp_im in zip(sha256s, fp_ims):
+  for pool_map, fp_im in zip(pool_maps, fp_ims):
     fpp_im = Path(fp_im)
     subdir = str(fpp_im.parent.relative_to(fp_in))
+    #subdir = '' if subdir is '.' else subdir
+    log.debug(subdir)
     
-
     if opt_identity:
       subdirs = subdir.split('/')
       if not len(subdirs) > 0:
@@ -120,27 +140,46 @@ def cli(ctx, opt_fp_in, opt_fp_out, opt_dataset, opt_data_store, opt_dir_media,
         log.error('exiting')
         return
       if opt_identity == 'subdir':
-        identity = subdirs[0]  # use first/only part
-      elif opt_identity == 'subdir_head':
-        identity = subdirs[0]  # use first part of subdir path
-      elif opt_identity == 'subdir_tail':
         identity = subdirs[-1]  # use last part of subdir path
+      elif opt_identity == 'numeric':
+        identity = indentity_count  # use incrementing number
+        indentity_count += 1
     else:
-      identity = indentity_count  # use incrementing number
-      indentity_count += 1
+      identity = ''
 
     data.append({
       'subdir': subdir,
+      'num_channels': pool_map['num_channels'],
       'fn': fpp_im.stem,
       'ext': fpp_im.suffix.replace('.',''),
-      'sha256': sha256,
+      'sha256': pool_map['sha256'],
       'uuid': uuid.uuid4(),
-      'identity_key': identity
+      'identity_key': identity,
+      'width': pool_map['width'],
+      'height': pool_map['height']
       })
 
+  # create dataframe
   df_records = pd.DataFrame.from_dict(data)
+
+  df_records.index.name = 'index'  # reassign 'index' as primary key column
+  # write to CSV
+  file_utils.mkdirs(fp_out)
+  df_records.to_csv(fp_out)
+  # done
+  log.info(f'wrote {len(df_records)} rows to "{fp_out}"')
+  # save script
+  cmd_line = ' '.join(sys.argv)
+  file_utils.write_text(cmd_line, '{}.sh'.format(fp_out))
+
+
+'''
+# create dataframe
+  df_records = pd.DataFrame.from_dict(data)
+
+  # add identity key (used for associating identity)
   if opt_identity:
-    log.info(f'adding identity index using: "{opt_identity}". This may take a while...')
+    log.info(f'adding identity index using: "{opt_identity}" subdirectory')
     # convert dict to DataFrame
     # sort based on identity_key
     df_records = df_records.sort_values(by=['identity_key'], ascending=True)
@@ -149,19 +188,17 @@ def cli(ctx, opt_fp_in, opt_fp_out, opt_dataset, opt_data_store, opt_dir_media,
     # populate the identity_index
     df_records_identity_groups = df_records.groupby('identity_key')
     # enumerate groups to create identity indices
+    log.info(f'updating records with identity_key. This may take a while...')
+    st = time.time()
     for identity_index, df_records_identity_group_tuple in enumerate(df_records_identity_groups):
       identity_key, df_records_identity_group = df_records_identity_group_tuple
       for ds_record in df_records_identity_group.itertuples():
         df_records.at[ds_record.Index, 'identity_index'] = identity_index
     # reset index after being sorted
     df_records = df_records.reset_index(drop=True)
+    log.debug('update time: {:.2f}s'.format(time.time() - st))
   else:
     # name everyone person 1, 2, 3...
+    df_records = df_records.sort_values(by=['subdir'], ascending=True)
     pass
-
-  df_records.index.name = 'index'  # reassign 'index' as primary key column
-  # write to CSV
-  file_utils.mkdirs(fp_out)
-  df_records.to_csv(fp_out)
-  # done
-  log.info(f'wrote rows: {len(df_records)} to {fp_out}')
\ No newline at end of file
+'''
\ No newline at end of file
diff --git a/megapixels/commands/datasets/msceleb.py b/megapixels/commands/datasets/msceleb.py
new file mode 100644
index 00000000..969a1df2
--- /dev/null
+++ b/megapixels/commands/datasets/msceleb.py
@@ -0,0 +1,66 @@
+'''
+Converts MsCelebV1-ImageThumbnails.part.00.tsv to names and images
+'''
+import click
+
+from app.settings import types
+from app.models.dataset import Dataset
+from app.utils import click_utils
+from app.settings import app_cfg as cfg
+from app.utils.logger_utils import Logger
+
+log = Logger.getLogger()
+
+@click.command()
+@click.option('-i', '--input', 'opt_fp_in', required=True,
+  help='Path to input TSV file')
+@click.option('-o', '--output', 'opt_fp_out', required=True,
+  help='Output path for images')
+@click.option('--slice', 'opt_slice', type=(int, int), default=(None, None),
+  help='Slice list of files')
+@click.pass_context
+def cli(ctx, opt_fp_in, opt_fp_out, opt_slice):
+  """Converts MSCeleb TSV to images"""
+  
+  import sys
+  import os
+  from glob import glob
+  from os.path import join
+  from pathlib import Path
+  import time
+  import base64
+  from io import BytesIO
+
+  import pandas as pd
+  import cv2 as cv
+  from PIL import Image
+  from tqdm import tqdm
+  
+  from app.utils import file_utils, im_utils
+  from app.models.data_store import DataStore
+
+
+  log = Logger.getLogger()
+  log.debug(f'opening "{opt_fp_in}" ...')
+  try:
+    n_lines = sum(1 for line in open(opt_fp_in))
+  except:
+    n_lines = 1
+    
+  log.debug('{:,}'.format(n_lines))
+
+  with open(opt_fp_in, 'rb') as fp:
+    for data_line in tqdm(fp, total=n_lines):
+      try:
+        freebase_mid, query_name, search_rank, url_image, url_page, b64_bytes = data_line.decode().split('\t')
+        # decode image
+        im64 = base64.b64decode(b64_bytes)
+        im = Image.open(BytesIO(im64))
+        # save image
+        dir_out = join(opt_fp_out, freebase_mid)
+        Path(dir_out).mkdir(parents=True, exist_ok=True)
+        idx = len(os.listdir(dir_out))
+        fp_out = join(dir_out, '{}.jpg'.format(file_utils.zpad(idx)))
+        im.save(fp_out, quality=100)
+      except Exception as e:
+        log.error('Could not process: {}, {}. Error: {}'.format(query_name, url_image, e))
diff --git a/megapixels/commands/datasets/msceleb_names.py b/megapixels/commands/datasets/msceleb_names.py
new file mode 100644
index 00000000..6ee2ad9a
--- /dev/null
+++ b/megapixels/commands/datasets/msceleb_names.py
@@ -0,0 +1,57 @@
+'''
+Converts MsCelebV1-ImageThumbnails.part.00.tsv to names and images
+'''
+import click
+
+from app.settings import types
+from app.models.dataset import Dataset
+from app.utils import click_utils
+from app.settings import app_cfg as cfg
+from app.utils.logger_utils import Logger
+
+log = Logger.getLogger()
+
+@click.command()
+@click.option('-i', '--input', 'opt_fp_in', required=True,
+  help='Path to input TSV file')
+@click.option('-o', '--output', 'opt_fp_out', required=True,
+  help='Output path for images')
+@click.pass_context
+def cli(ctx, opt_fp_in, opt_fp_out):
+  """Converts MSCeleb TSV to names file with image count"""
+  
+  import sys
+  import os
+  from glob import glob
+  from os.path import join
+  from pathlib import Path
+  import time
+  import base64
+  from io import BytesIO
+
+  import pandas as pd
+  import cv2 as cv
+  from PIL import Image
+  from tqdm import tqdm
+  
+  from app.utils import file_utils, im_utils
+  from app.models.data_store import DataStore
+
+
+  log = Logger.getLogger()
+  log.debug(f'opening "{opt_fp_in}" ...')
+  n_lines = sum(1 for line in open(opt_fp_in))
+  log.debug('{:,}'.format(n_lines))
+
+  with open(opt_fp_in, 'rb') as fp:
+    for data_line in tqdm(fp, total=n_lines):
+      freebase_mid, query_name, search_rank, url_image, url_page, b64_bytes = data_line.decode().split('\t')
+      # decode image
+      im64 = base64.b64decode(b64_bytes)
+      im = Image.open(BytesIO(im64))
+      # save image
+      dir_out = join(opt_fp_out, freebase_mid)
+      Path(dir_out).mkdir(parents=True, exist_ok=True)
+      idx = len(os.listdir(dir_out))
+      fp_out = join(dir_out, '{}.jpg'.format(file_utils.zpad(idx)))
+      im.save(fp_out, quality=100)
diff --git a/megapixels/commands/demo/face_search.py b/megapixels/commands/demo/face_search.py
index 6e4bcdad..ca0b8016 100644
--- a/megapixels/commands/demo/face_search.py
+++ b/megapixels/commands/demo/face_search.py
@@ -15,7 +15,7 @@ log = Logger.getLogger()
   type=cfg.DataStoreVar,
   default=click_utils.get_default(types.DataStore.SSD),
   show_default=True,
-  help=click_utils.show_help(types.Dataset))
+  help=click_utils.show_help(types.DataStore))
 @click.option('--dataset', 'opt_dataset',
   type=cfg.DatasetVar,
   required=True,
@@ -26,7 +26,7 @@ log = Logger.getLogger()
 @click.option('--gpu', 'opt_gpu', default=0, 
   help='GPU index (use -1 for CPU')
 @click.pass_context
-def cli(ctx, opt_fp_in, opt_data_store, opt_dataset, opt_gpu):
+def cli(ctx, opt_fp_in, opt_data_store, opt_dataset, opt_results, opt_gpu):
   """Display image info"""
   
   import sys