adding verbiage and timing

3 files changed, 39 insertions, 27 deletions

diff --git a/megapixels/app/models/sql_factory.py b/megapixels/app/models/sql_factory.py
index 9a44941b..02b722df 100644
--- a/megapixels/app/models/sql_factory.py
+++ b/megapixels/app/models/sql_factory.py
@@ -98,7 +98,7 @@ class SqlDataset:
       return None
     session = Session()
     # for obj in session.query(table).filter_by(id=id):
-    print(table)
+    # print(table)
     obj = session.query(table).filter(table.id == id).first()
     session.close()
     return obj.toJSON()
diff --git a/megapixels/app/server/api.py b/megapixels/app/server/api.py
index bc60118c..b3447eb1 100644
--- a/megapixels/app/server/api.py
+++ b/megapixels/app/server/api.py
@@ -15,24 +15,32 @@ from app.utils.im_utils import pil2np
 sanitize_re = re.compile('[\W]+')
 valid_exts = ['.gif', '.jpg', '.jpeg', '.png']
 
+LIMIT = 9
+THRESHOLD = 0.3
+
 api = Blueprint('api', __name__)
 
 faiss_datasets = load_faiss_databases()
 
 @api.route('/')
 def index():
+  """List the datasets and their fields"""
   return jsonify({ 'datasets': list_datasets() })
 
+
 @api.route('/dataset/<name>')
 def show(name):
+  """Show the data that a dataset will return"""
   dataset = get_dataset(name)
   if dataset:
     return jsonify(dataset.describe())
   else:
     return jsonify({ 'status': 404 })
 
+
 @api.route('/dataset/<name>/face/', methods=['POST'])
 def upload(name):
+  """Query an image against FAISS and return the matching identities"""
   start = time.time()
   dataset = get_dataset(name)
   if name not in faiss_datasets:
@@ -52,31 +60,39 @@ def upload(name):
 
   im = Image.open(file.stream).convert('RGB')
   im_np = pil2np(im)
-  
+
   # Face detection
   detector = face_detector.DetectorDLIBHOG()
 
   # get detection as BBox object
   bboxes = detector.detect(im_np, largest=True)
-  if not len(bboxes):
+  if not bboxes or not len(bboxes):
     return jsonify({
       'error': 'bbox'
     })
   bbox = bboxes[0]
+  if not bbox:
+    return jsonify({
+      'error': 'bbox'
+    })
+
   dim = im_np.shape[:2][::-1]
   bbox = bbox.to_dim(dim)  # convert back to real dimensions
+  print("got bbox")
+  if not bbox:
+    return jsonify({
+      'error': 'bbox'
+    })
 
-  # face recognition/vector
+  # extract 128-D vector
   recognition = face_recognition.RecognitionDLIB(gpu=-1)
   vec = recognition.vec(im_np, bbox)
-
-  # print(vec)
   query = np.array([ vec ]).astype('float32')
   
-  # query FAISS!
-  distances, indexes = faiss_dataset.search(query, 10)
+  # query FAISS
+  distances, indexes = faiss_dataset.search(query, LIMIT)
 
-  if len(indexes) == 0:
+  if len(indexes) == 0 or len(indexes[0]) == 0:
     return jsonify({
       'error': 'nomatch'
     })
@@ -85,36 +101,32 @@ def upload(name):
   distances = distances[0]
   indexes = indexes[0]
 
-  if len(indexes) == 0:
-    return jsonify({
-      'error': 'nomatch'
-    })
-
-  lookup = {}
-  ids = [i+1 for i in indexes]
+  dists = []
+  ids = []
   for _d, _i in zip(distances, indexes):
-    lookup[_i+1] = _d
+    if _d <= THRESHOLD:
+      dists.append(round(float(_d), 2))
+      ids.append(_i+1)
 
-  print(distances)
-  print(indexes)
+  results = [ dataset.get_identity(_i) for _i in ids ]
 
-  # with the result we have an ID
-  # query the sql dataset for the UUID etc here
+  print(distances)
+  print(ids)
 
   query = {
-    'timing': time.time() - start,
+    'timing': round(time.time() - start, 3),
   }
-  results = [ dataset.get_identity(id) for id in ids ]
-
   print(results)
   return jsonify({
+    'query': query,
     'results': results,
-    'distances': distances.tolist(),
-    'indexes': indexes.tolist(),
+    'distances': dists,
   })
 
+
 @api.route('/dataset/<name>/name', methods=['GET'])
 def name_lookup(dataset):
+  """Find a name in the dataset"""
   start = time.time()
   dataset = get_dataset(name)
 
diff --git a/megapixels/app/site/parser.py b/megapixels/app/site/parser.py
index ecfae0cb..b3d3a8c2 100644
--- a/megapixels/app/site/parser.py
+++ b/megapixels/app/site/parser.py
@@ -64,7 +64,7 @@ def format_applet(section, s3_path):
   else:
     command = payload[0]
     opt = None
-  if command == 'python':
+  if command == 'python' or command == 'javascript' or command == 'code':
     return format_section([ section ], s3_path)
 
   applet['command'] = command