initial version

app.py

@@ -0,0 +1,75 @@
+import os
+from pathlib import Path
+from subprocess import run
+
+import gradio as gr
+
+
+yolo_input_size = 384
+versions = ('2_v108', '4_v109', '0_int6', '1_v110', '3_v111')
+
+score_thr = 0.025
+zoom_score_thr = 0.35
+iou_thr = 0.6
+max_det = 1
+yolo_ens = 'fast'  # fast, val, detect, detect_internal, all
+output_size = (512, 512)
+bs = 1 #128 if 'CUDA_VERSION' in os.environ else 16
+
+project_dir = None
+working = Path(os.getcwd())
+modelbox = working / 'models'
+checkpoint_files = [modelbox / f'yolov5_l6_{yolo_input_size}_fold{x}.pt' for x in versions]
+image_root = working / 'images' / 'subdir'
+
+image_urls = [
+    # Negatives
+    'https://upload.wikimedia.org/wikipedia/commons/c/c5/Common_Dolphin.jpg',
+    'https://upload.wikimedia.org/wikipedia/commons/b/b8/Beluga847.jpg',
+    'https://upload.wikimedia.org/wikipedia/commons/e/ea/Beluga_1_1999-07-03.jpg',
+    'https://upload.wikimedia.org/wikipedia/commons/2/2b/Whale_Watching_in_Gloucester%2C_Massachusetts_5.jpg',
+
+    # Positives
+    '/kaggle/input/happy-whale-and-dolphin/test_images/00098d1376dab2.jpg',
+]
+yolo_source = f'{image_root}/negative001.jpg'
+
+
+def pred_fn(image, fake=True):
+    if fake:
+        x0, x1 = (int(f * image.shape[0]) for f in (0.2, 0.8))
+        y0, y1 = (int(f * image.shape[1]) for f in (0.2, 0.8))
+        cropped_image = image[x0:x1, y0:y1, :]
+        response_str = f"This looks like a common dolphin, but I have not seen this individual before (0.834 confidence).\n" \
+                       "Go submit your photo on www.happywhale.com!"
+        return cropped_image, response_str
+
+    cropped_image, species = fast_yolo_crop(image)
+    test_embedding = get_test_embedding(embed_models, sizes)
+    
+    cosine_similarity = np.dot(comp_embeddings, test_embedding[0]) / n_models
+    cosine_distances = 1 - cosine_similarity
+    normalized_distances = cosine_distances / max_distance
+    normalized_similarities = 1 - normalized_distances
+
+    min_similarity = normalized_similarities.min()
+    max_similarity = normalized_similarities.max()
+    confidence = get_confidence(max_similarity, threshold)
+
+    print(f"Similarities: {min_similarity:.4f} ... {max_similarity:.4f}")
+    print(f"Threshold:", threshold)
+    
+    if max_similarity > threshold:
+        response_str = f"This looks like a {species} I have seen before ({confidence:.3f} confidence).\n" \
+                       "You might find its previous encounters on www.happywhale.com"
+    else:
+        response_str = f"This looks like a {species}, but I have not seen this individual before ({confidence:.3f} confidence).\n" \
+                       "Go submit your photo on www.happywhale.com!"
+
+    return cropped_image, response_str
+
+examples = [str(image_root / f'negative{i:03d}') for i in range(3)]
+
+demo = gr.Interface(fn=pred_fn, inputs="image", outputs=["image", "text"],
+                    examples=examples)
+demo.launch()

images/LICENSE.txt

@@ -0,0 +1,7 @@
+Images in this folder are licensed under the Creative Commons Attribution-Share Alike 4.0 International license:
+https://creativecommons.org/licenses/by-sa/4.0/deed.en
+
+Attribution/Origin:
+negative001.jpg https://upload.wikimedia.org/wikipedia/commons/2/2b/Whale_Watching_in_Gloucester%2C_Massachusetts_5.jpg
+negative002.jpg https://upload.wikimedia.org/wikipedia/commons/c/c5/Common_Dolphin.jpg
+