add compare esnet moodel

check_arch.py

@@ -0,0 +1,21 @@
+import h5py
+
+def print_model_details(file_path):
+    with h5py.File(file_path, 'r') as f:
+        print(f.keys())  # Print layers
+        print(len(f.keys())) 
+        print("")
+        for key in f.keys():
+            print(len(list(f[key].keys())))
+            print(f"{key}: {list(f[key].keys())}")  # Print details of each layer)
+print('rock')
+print_model_details('model_classification/rock-170.h5')
+print('mummified-170')
+print_model_details('model_classification/mummified-170.h5')
+print('BEiT')
+print_model_details('model_classification/fossil-142.h5')
+print('BEiT New')
+print_model_details('model_classification/fossil-new.h5')
+print("Newest:")
+print_model_details('model_classification/fossil-model.h5')
+

inference_resnet.py

@@ -16,7 +16,7 @@ import matplotlib.pyplot as plt
 from typing import Tuple
 from huggingface_hub import snapshot_download
 from labels import lookup_170
-import numpy as np 
+import numpy as np
 
 if not os.path.exists('model_classification'):
 
@@ -173,7 +173,11 @@ def inference_resnet_finer(x,model,size=576,n_classes=170,n_top=10):
     
     cropped = _clever_crop(x,(size,size))[0]
     prep = preprocess(cropped,size=size)
-    logits = tf.nn.softmax(model.predict(np.array([prep]))[1][0]).cpu().numpy()
+    logits = model.predict(np.array([prep]))
+    print(logits)
+    logits = tf.nn.softmax(logits[1][0]).cpu().numpy()
+    print(logits)
     top_n = select_top_n(logits,n=n_top)
+    print(top_n)
     
     return parse_results(top_n,logits)

inference_resnet_v2.py

@@ -0,0 +1,79 @@
+import tensorflow as tf
+
+tf.config.set_visible_devices([], 'GPU')
+# gpu_devices = tf.config.experimental.list_physical_devices('GPU')
+# if gpu_devices:
+#     tf.config.experimental.set_memory_growth(gpu_devices[0], True)
+# else:
+#     print(f"TensorFlow device: {gpu_devices}")
+
+from keras.applications import resnet
+import tensorflow as tf
+import keras
+import tensorflow.keras.layers as L
+import os
+
+from tensorflow.keras.layers import Dense, GlobalAveragePooling2D
+import matplotlib.pyplot as plt
+from typing import Tuple
+from huggingface_hub import snapshot_download
+from labels import lookup_140
+import numpy as np 
+
+if not os.path.exists('model_classification'):
+
+    REPO_ID='Serrelab/fossil_classification_models'
+    token = os.getenv('READ_TOKEN')
+    print('read token:',token)
+    if token is None:
+        print("warning! A read token in env variables is needed for authentication.")
+    snapshot_download(repo_id=REPO_ID,token=token,repo_type='model',local_dir='model_classification')
+
+def  get_resnet_model(model_path):
+    cce = tf.keras.losses.categorical_crossentropy
+    model = keras.models.load_model(model_path, custom_objects = {"cce":cce})
+    g = keras.Model(model.input, model.layers[2].output)
+    # out = tf.keras.layers.Activation('relu')(g_.output)
+    # g = tf.keras.Model(model.input, out)
+    h = keras.Model(model.layers[3].input, model.layers[-1].output)
+    return model, g, h
+
+
+def select_top_n(preds,n=10):
+    top_n = np.argsort(preds)[-n:][::-1]
+    return top_n
+
+
+def parse_results(top_n,logits):
+    results = {}
+    for n in top_n:
+        label = lookup_140[n]
+        results[label] = float(logits[n])
+    return results 
+
+def inference_resnet_embedding(x,model,size=384,n_classes=140,n_top=10):
+        
+        
+    x = tf.image.resize(x, (size, size))
+    x = tf.reshape(x, (-1, 384, 384, 3))/255
+    embedding = model.predict(np.array([x]))[0][0]
+   
+    
+    return embedding 
+
+def inference_resnet_finer(x,model,size=384,n_classes=142,n_top=10):
+    
+    x = tf.image.resize(x, (size, size))
+    x = tf.reshape(x, (-1, 384, 384, 3))/255
+    # _, batch_logits = model.predict(x)
+    # predictions = tf.math.top_k(batch_logits, k=10)
+    # print(predictions)
+    logits = model.predict(np.array([x]))
+    print(logits)
+    logits = tf.nn.softmax(logits[1][0]).cpu().numpy()
+    print(logits)
+    top_n = select_top_n(logits,n=n_top)
+    print(top_n)
+    
+    return parse_results(top_n,logits)
+