added FR demo notebook

6 files changed, 225 insertions, 132 deletions

diff --git a/megapixels/commands/cv/faces_to_csv.py b/megapixels/commands/cv/_old_files_to_face_rois.py
index 1fd47571..d92cbd74 100644
--- a/megapixels/commands/cv/faces_to_csv.py
+++ b/megapixels/commands/cv/_old_files_to_face_rois.py
@@ -12,8 +12,8 @@ from app.settings import app_cfg as cfg
 color_filters = {'color': 1, 'gray': 2, 'all': 3}
 
 @click.command()
-@click.option('-i', '--input', 'opt_dirs_in', required=True, multiple=True,
-  help='Input directory')
+@click.option('-i', '--input', 'opt_fp_files', required=True,
+  help='Input file meta CSV')
 @click.option('-o', '--output', 'opt_fp_out', required=True,
   help='Output CSV')
 @click.option('-e', '--ext', 'opt_ext',
diff --git a/megapixels/commands/cv/embeddings.py b/megapixels/commands/cv/embeddings.py
deleted file mode 100644
index 9cb26ae7..00000000
--- a/megapixels/commands/cv/embeddings.py
+++ /dev/null
@@ -1,100 +0,0 @@
-"""
-Crop images to prepare for training
-"""
-
-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', required=True,
-  help='Input directory')
-@click.option('-r', '--records', 'opt_fp_records', required=True,
-  help='Input directory')
-@click.option('-m', '--media', 'opt_fp_media', required=True,
-  help='Image directory')
-@click.option('-o', '--output', 'opt_fp_out', required=True,
-  help='Output CSV')
-@click.option('--size', 'opt_size', 
-  type=(int, int), default=(300, 300),
-  help='Output image size')
-@click.option('-g', '--gpu', 'opt_gpu', default=0,
-  help='GPU index')
-@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('-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.pass_context
-def cli(ctx, opt_fp_in, opt_fp_records, opt_fp_out, opt_fp_media, opt_size, opt_gpu,
-  opt_slice, opt_jitters, opt_padding, opt_force):
-  """Converts frames with faces to CSV of rows"""
-  
-  import sys
-  import os
-  from os.path import join
-  from pathlib import Path
-  
-  from tqdm import tqdm
-  import numpy as np
-  import dlib  # must keep a local reference for dlib
-  import cv2 as cv
-  import dlib
-  import pandas as pd
-
-  from app.utils import logger_utils, file_utils, im_utils
-  from app.models.bbox import BBox
-  from app.processors import face_recognition
-
-  # -------------------------------------------------
-  # init here
-
-  log = logger_utils.Logger.getLogger()
-
-  if not opt_force and Path(opt_fp_out).exists():
-    log.error('File exists. Use "-f / --force" to overwite')
-    return
-  
-  # init dlib FR
-  facerec = face_recognition.RecognitionDLIB()
-
-  # load data
-  df_rois = pd.read_csv(opt_fp_in)
-  df_records = pd.read_csv(opt_fp_records)
-
-  if opt_slice:
-    df_rois = df_rois[opt_slice[0]:opt_slice[1]]
-  log.info('Processing {:,} rows'.format(len(df_rois)))
-  nrows = len(df_rois)
-
-  # face vecs
-  vecs = []
-
-  for roi_idx, row in tqdm(df_rois.iterrows(), total=nrows):
-    # make image path
-    record_id = int(row['id'])
-    df = df_records.iloc[record_id]
-    fp_im = join(opt_fp_media, df['subdir'], '{}.{}'.format(df['fn'], df['ext'])) 
-    # load image
-    im = cv.imread(fp_im)
-    # make bbox
-    xywh = [row['x'], row['y'], row['w'] , row['h']]
-    bbox = BBox.from_xywh(*xywh)
-    # scale to actual image size
-    dim = (row['image_width'], row['image_height'])
-    bbox_dim = bbox.to_dim(dim)
-    # compute vec
-    vec = facerec.vec(im, bbox_dim, jitters=opt_jitters, padding=opt_padding)
-    vec_str = ','.join([repr(x) for x in vec])
-    vecs.append( {'id': row['id'], 'vec': vec_str})
-  
-  # save data
-  file_utils.mkdirs(opt_fp_out)
-  df_vecs = pd.DataFrame.from_dict(vecs)
-  df_vecs.to_csv(opt_fp_out, index=False)
-  log.info('saved {:,} lines to {}'.format(len(df_vecs), opt_fp_out))
\ No newline at end of file
diff --git a/megapixels/commands/cv/face_vec_to_csv.py b/megapixels/commands/cv/face_vec_to_csv.py
new file mode 100644
index 00000000..6c9fad09
--- /dev/null
+++ b/megapixels/commands/cv/face_vec_to_csv.py
@@ -0,0 +1,110 @@
+"""
+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('-i', '--input', 'opt_fp_files', required=True,
+  help='Input ROI CSV')
+@click.option('-r', '--rois', 'opt_fp_rois', required=True,
+  help='Input ROI CSV')
+@click.option('-m', '--media', 'opt_dir_media', required=True,
+  help='Input media directory')
+@click.option('-o', '--output', 'opt_fp_out', required=True,
+  help='Output CSV')
+@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_files, opt_fp_rois, opt_dir_media, opt_fp_out, 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.utils import logger_utils, file_utils, im_utils
+  from app.processors import face_recognition
+
+
+  # -------------------------------------------------
+  # init here
+
+  log = logger_utils.Logger.getLogger()
+
+  # init face processors
+  facerec = face_recognition.RecognitionDLIB()
+
+  # load data
+  df_file_meta = pd.read_csv(opt_fp_files)
+  df_rois = pd.read_csv(opt_fp_rois)
+
+  if not opt_force and Path(opt_fp_out).exists():
+    log.error('File exists. Use "-f / --force" to overwite')
+    return
+  
+  if opt_slice:
+    df_rois = df_rois[opt_slice[0]:opt_slice[1]]
+  
+  # -------------------------------------------------
+  # process here
+
+  df_img_groups = df_rois.groupby('image_index')
+  log.debug('processing {:,} groups'.format(len(df_img_groups)))
+
+  vecs = []
+
+  for image_index, df_img_group in tqdm(df_img_groups):
+    # make fp
+    roi_index = df_img_group.index.values[0]
+    file_meta = df_file_meta.iloc[image_index]  # locate image meta
+    fp_im = join(opt_dir_media, file_meta.subdir, '{}.{}'.format(file_meta.fn, file_meta.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, 'image_index': image_index, 'vec': vec_str})
+
+
+  # save date
+  file_utils.mkdirs(opt_fp_out)
+  df = pd.DataFrame.from_dict(vecs)
+  df.index.name = 'index'
+  df.to_csv(opt_fp_out)
\ No newline at end of file
diff --git a/megapixels/commands/cv/faces_to_csv_indexed.py b/megapixels/commands/cv/files_to_rois.py
index ef958f89..1aaf991c 100644
--- a/megapixels/commands/cv/faces_to_csv_indexed.py
+++ b/megapixels/commands/cv/files_to_rois.py
@@ -100,7 +100,7 @@ def cli(ctx, opt_fp_in, opt_dir_media, opt_fp_out, opt_size, opt_detector_type,
   data = []
 
   for df_file in tqdm(df_files.itertuples(), total=len(df_files)):
-    fp_im = join(opt_dir_media, df_file.subdir, '{}.{}'.format(df_file.fn, df_file.ext))
+    fp_im = join(opt_dir_media, str(df_file.subdir), f'{df_file.fn}.{df_file.ext}')
     im = cv.imread(fp_im)
 
     # filter out color or grayscale iamges
@@ -115,7 +115,7 @@ def cli(ctx, opt_fp_in, opt_dir_media, opt_fp_out, opt_size, opt_detector_type,
         continue
         
     try:
-      bboxes = detector.detect(im, opt_size=opt_size, opt_pyramids=opt_pyramids, opt_largest=opt_largest)
+      bboxes = detector.detect(im, size=opt_size, pyramids=opt_pyramids, largest=opt_largest)
     except Exception as e:
       log.error('could not detect: {}'.format(fp_im))
       log.error('{}'.format(e))
diff --git a/megapixels/commands/datasets/file_meta.py b/megapixels/commands/datasets/file_meta.py
new file mode 100644
index 00000000..e1456f44
--- /dev/null
+++ b/megapixels/commands/datasets/file_meta.py
@@ -0,0 +1,84 @@
+"""
+Begin with this file to process folder of images
+- Converts folders and subdirectories into CSV with file attributes split
+"""
+import click
+
+from app.settings import types
+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='Input directory')
+@click.option('-o', '--output', 'opt_fp_out', required=True,
+  help='Output file for file meta CSV')
+@click.option('--slice', 'opt_slice', type=(int, int), default=(None, None),
+  help='Slice list of files')
+@click.option('--recursive/--no-recursive', 'opt_recursive', is_flag=True, default=False,
+  help='Use glob recursion (slower)')
+@click.option('-t', '--threads', 'opt_threads', default=4,
+  help='Number of threads')
+@click.option('-f', '--force', 'opt_force', is_flag=True,
+  help='Force overwrite file')
+@click.pass_context
+def cli(ctx, opt_fp_in, opt_fp_out, opt_slice, opt_recursive, opt_threads, opt_force):
+  """Multithreading test"""
+  
+  from glob import glob
+  from os.path import join
+  from pathlib import Path
+  import time
+  from multiprocessing.dummy import Pool as ThreadPool 
+  import random
+
+  import pandas as pd
+  from tqdm import tqdm
+  from glob import glob
+  
+  from app.utils import file_utils, im_utils
+
+
+  if not opt_force and Path(opt_fp_out).exists():
+    log.error('File exists. Use "-f / --force" to overwite')
+    return
+
+  fp_ims = []
+  log.info(f'Globbing {opt_fp_in}')
+  for ext in ['jpg', 'png']:
+    if opt_recursive:
+      fp_glob = join(opt_fp_in, '**/*.{}'.format(ext))
+      fp_ims += glob(fp_glob, recursive=True)
+    else:
+      fp_glob = join(opt_fp_in, '*.{}'.format(ext))
+      fp_ims += glob(fp_glob)
+
+  if not fp_ims:
+    log.warn('No images. Try with "--recursive"')
+    return
+
+  if opt_slice:
+    fp_ims = fp_ims[opt_slice[0]:opt_slice[1]]
+
+  log.info('Processing {:,} images'.format(len(fp_ims)))
+
+
+  # convert data to dict
+  data = []
+  for i, fp_im in enumerate(tqdm(fp_ims)):
+    fpp_im = Path(fp_im)
+    subdir = str(fpp_im.parent.relative_to(opt_fp_in))
+    data.append( {
+      'subdir': subdir,
+      'fn': fpp_im.stem,
+      'ext': fpp_im.suffix.replace('.','')
+      })
+
+  # save to CSV
+  file_utils.mkdirs(opt_fp_out)
+  df = pd.DataFrame.from_dict(data)
+  df.index.name = 'index'
+  df.to_csv(opt_fp_out)
\ No newline at end of file
diff --git a/megapixels/commands/datasets/sha256.py b/megapixels/commands/datasets/sha256.py
index c04fb504..4c734073 100644
--- a/megapixels/commands/datasets/sha256.py
+++ b/megapixels/commands/datasets/sha256.py
@@ -10,18 +10,18 @@ log = Logger.getLogger()
 @click.command()
 @click.option('-i', '--input', 'opt_fp_in', required=True,
   help='Input directory')
-@click.option('-o', '--output', 'opt_fp_out',
+@click.option('-m', '--media', 'opt_dir_media', required=True,
+  help='Input media directory')
+@click.option('-o', '--output', 'opt_fp_out', required=True,
   help='Output directory')
 @click.option('--slice', 'opt_slice', type=(int, int), default=(None, None),
   help='Slice list of files')
-@click.option('--recursive/--no-recursive', 'opt_recursive', is_flag=True, default=False,
-  help='Use glob recursion (slower)')
 @click.option('-t', '--threads', 'opt_threads', default=4,
   help='Number of threads')
 @click.option('-f', '--force', 'opt_force', is_flag=True,
   help='Force overwrite file')
 @click.pass_context
-def cli(ctx, opt_fp_in, opt_fp_out, opt_slice, opt_recursive, opt_threads, opt_force):
+def cli(ctx, opt_fp_in, opt_dir_media, opt_fp_out, opt_slice, opt_threads, opt_force):
   """Multithreading test"""
   
   from glob import glob
@@ -42,47 +42,46 @@ def cli(ctx, opt_fp_in, opt_fp_out, opt_slice, opt_recursive, opt_threads, opt_f
     log.error('File exists. Use "-f / --force" to overwite')
     return
 
-  fp_ims = []
-  for ext in ['jpg', 'png']:
-    if opt_recursive:
-      fp_glob = join(opt_fp_in, '**/*.{}'.format(ext))
-      fp_ims += glob(fp_glob, recursive=True)
-    else:
-      fp_glob = join(opt_fp_in, '*.{}'.format(ext))
-      fp_ims += glob(fp_glob)
+  df_files = pd.read_csv(opt_fp_in).set_index('index')
 
   if opt_slice:
-    fp_ims = fp_ims[opt_slice[0]:opt_slice[1]]
+    df_files = df_files[opt_slice[0]:opt_slice[1]]
 
-  log.info('Processing {:,} images'.format(len(fp_ims)))
+  log.info('Processing {:,} images'.format(len(df_files)))
 
-  pbar = tqdm(total=100)
+  
+  # prepare list of images to multithread into sha256s
+  file_objs = []
+  for ds_file in df_files.itertuples():
+    fp_im = join(opt_dir_media, str(ds_file.subdir), f"{ds_file.fn}.{ds_file.ext}")
+    file_objs.append({'fp': fp_im, 'index': ds_file.Index})
+
+  # convert to thread pool
+  pbar = tqdm(total=len(file_objs))
 
-  def as_sha256(fp_im):
+  def as_sha256(file_obj):
     pbar.update(1)
-    return file_utils.sha256(fp_im)
+    file_obj['sha256'] = file_utils.sha256(file_obj['fp'])
+    return file_obj
 
   # multithread pool
+  pool_file_objs = []
   st = time.time()
   pool = ThreadPool(opt_threads) 
-  with tqdm(total=len(fp_ims)) as pbar:
-    sha256s = pool.map(as_sha256, fp_ims)
+  with tqdm(total=len(file_objs)) as pbar:
+    pool_file_objs = pool.map(as_sha256, file_objs)
   pbar.close()
-
+  
   # convert data to dict
   data = []
-  for i, fp_im in enumerate(fp_ims):
-    fpp_im = Path(fp_im)
-    subdir = str(fpp_im.parent.relative_to(opt_fp_in))
-    sha256 = sha256s[i]
+  for pool_file_obj in pool_file_objs:
     data.append( {
-      'sha256': sha256, 
-      'subdir': subdir,
-      'fn': fpp_im.stem,
-      'ext': fpp_im.suffix.replace('.','')
+      'sha256': pool_file_obj['sha256'], 
+      'index': pool_file_obj['index']
       })
 
   # save to CSV
+  file_utils.mkdirs(opt_fp_out)
   df = pd.DataFrame.from_dict(data)
   df.to_csv(opt_fp_out, index=False)