Модель TFLite выдает разные результаты в Android, чем в приложении Kivy. ⇐ Android

[Anonymous]

Я столкнулся с проблемой, которую не могу решить: моя модель tflite выдает разные данные, когда я запускаю ее на Python или Android, что приводит к ошибкам и сбою на Android. Я разрабатываю приложение Kivy с tflite, которое обнаруживает шахматные фигуры на шахматной доске. Я конвертировал свою модель YOLOv8 в формат tflite. Проблема возникает при выполнении вывода изображения в этом методе:
def get_positions(model_path, img_name):
print("Loading model...")
interpreter = Interpreter(model_path=model_path)
interpreter.allocate_tensors()

input_details = interpreter.get_input_details()
output_details = interpreter.get_output_details()
print("Input details:", input_details)
print("Output details:", output_details)

print("Loading image...")
image = cv2.imread(img_name)
if image is None:
raise ValueError(f"Could not load the image from {img_name}")

print("Original image dimensions:", image.shape)
image_height = input_details[0]['shape'][1]
image_width = input_details[0]['shape'][2]
resized_image = cv2.resize(image, (image_width, image_height))
input_image = np.array(resized_image, dtype=np.float32) / 255.0
input_image = input_image[np.newaxis, :]
print("Resized image dimensions:", input_image.shape)

interpreter.set_tensor(input_details[0]['index'], input_image)
interpreter.invoke()

output = interpreter.get_tensor(output_details[0]['index'])
output = output[0]
output = output.T
print("Model output:", output)

boxes_xywh = output[..., :4]
scores = np.max(output[..., 4:], axis=1)
print("Scores:", scores)
classes = np.argmax(output[..., 4:], axis=1)

indices = cv2.dnn.NMSBoxes(boxes_xywh.tolist(), scores.tolist(), confidence_threshold, iou_threshold)
print("Indices after NMS:", indices)

results = []
for i in indices:
if scores >= confidence_threshold:
x_center, y_center, width, height = boxes_xywh
x_center, width = x_center * image_width, width * image_width
y_center, height = y_center * image_height, height * image_height
x_center /= image_width
y_center /= image_height
width /= image_width
height /= image_height

result = {
'class_id': classes,
'class_name': CLASSES[classes],
'x_center': x_center,
'y_center': y_center,
'width': width,
'height': height,
'confidence': scores
}
print("Result:", result)
results.append(result)

return results

На моем компьютере все работает как положено, но когда я создаю APK на Android, происходит сбой, поскольку соответствующие данные не генерируются.
Вот журналы отладки с моего компьютера (работает хорошо)
\`Loading model... INFO: Created TensorFlow Lite XNNPACK delegate for CPU. Input details: \[{'name': 'inputs_0', 'index': 0, 'shape': array(\[ 1, 640, 640, 3\], dtype=int32), 'shape_signature': array(\[ 1, 640, 640, 3\], dtype=int32), 'dtype': \, 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array(\[\], dtype=float32), 'zero_points': array(\[\], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}\] Output details: \[{'name': 'Identity', 'index': 409, 'shape': array(\[ 1, 17, 8400\], dtype=int32), 'shape_signature': array(\[ 1, 17, 8400\], dtype=int32), 'dtype': \, 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array(\[\], dtype=float32), 'zero_points': array(\[\], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}\] Loading image... Original image dimensions: (1181, 1690, 3) Resized image dimensions: (1, 640, 640, 3) Model output: \[\[1.1006817e-02 3.4101062e-02 1.9614611e-02 ... 2.6802209e-05 3.2099266e-05 2.3671286e-05\] \[3.3980943e-02 1.1495238e-02 5.1065966e-02 ... 3.0108062e-05 2.4680974e-05 3.4402736e-05\] \[4.6541955e-02 7.9353973e-03 6.3635595e-02 ... 1.8403694e-05 4.7014169e-06 2.9310922e-05\] ... \[8.8734090e-01 9.7479749e-01 1.9226933e-01 ... 2.3383725e-06 1.9063839e-06 8.1505220e-07\] \[9.3079507e-01 9.7354954e-01 1.9993711e-01 ... 2.2026159e-06 2.1150520e-06 8.2314017e-07\] \[9.7550833e-01 9.7113740e-01 1.7806405e-01 ... 1.6881496e-06 1.1987435e-06 6.0122733e-07\]\] Scores: \[4.7006957e-05 8.5894673e-05 4.0709678e-05 ... 3.7478815e-06 3.3213523e-06 2.8609916e-06\] Indices after NMS: \[7601 7853 7668 6563 7878 7088 7262 7287 6536 2125 6607 3547 6808 5907 7688 1336 6896 6972 7068 2867 4295 7497 5086 448 4266 5798 705 7077 6478 6483 7242\] Result: {'class_id': 7, 'class_name': 'white bishop', 'x_center': 0.05852939561009407, 'y_center': 0.773659348487854, 'width': 0.08957379311323166, 'height': 0.14608508348464966, 'confidence': 0.92357063} Result: {'class_id': 12, 'class_name': 'white rook', 'x_center': 0.3321531414985657, 'y_center': 0.8937365412712097, 'width': 0.0860576331615448, 'height': 0.13431024551391602, 'confidence': 0.9208999} Result: {'class_id': 3, 'class_name': 'black knight', 'x_center': 0.7246953248977661, 'y_center': 0.7933675646781921, 'width': 0.0967983603477478, 'height': 0.13826501369476318, 'confidence': 0.9208062} .......... .......... \`

And from my Android app (it crashes)\`

`Loading model... 05-03 16:37:36.598 6567 30025 I python : Input details: [{'name': 'inputs_0', 'index': 0, 'shape': array([ 1, 640, 640, 3], dtype=int32), 'shape_signature': array([ 1, 640, 640, 3], dtype=int32), 'dtype': , 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}] 05-03 16:37:36.599 6567 30025 I python : Output details: [{'name': 'Identity', 'index': 409, 'shape': array([ 1, 17, 8400], dtype=int32), 'shape_signature': array([ 1, 17, 8400], dtype=int32), 'dtype': , 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}] 05-03 16:37:36.599 6567 30025 I python : Loading image... 05-03 16:37:36.654 6567 30025 I python : Original image dimensions: (1181, 1690, 3) 05-03 16:37:36.667 6567 30025 I python : Resized image dimensions: (1, 640, 640, 3) 05-03 16:37:39.163 6567 30025 I python : Model output: [[1.1006776e-02 3.4101110e-02 1.9614521e-02 ... 2.6802416e-05 05-03 16:37:39.163 6567 30025 I python : 3.2099608e-05 2.3671355e-05] 05-03 16:37:39.163 6567 30025 I python : [3.3980943e-02 1.1495264e-02 5.1065952e-02 ... 3.0108265e-05 05-03 16:37:39.163 6567 30025 I python : 2.4681047e-05 3.4402743e-05] 05-03 16:37:39.163 6567 30025 I python : [4.6541952e-02 7.9354066e-03 6.3635573e-02 ... 1.8403680e-05 05-03 16:37:39.163 6567 30025 I python : 4.7014123e-06 2.9310897e-05] 05-03 16:37:39.163 6567 30025 I python : ... 05-03 16:37:39.163 6567 30025 I python : [8.8734090e-01 9.7479749e-01 1.9226938e-01 ... 2.3383748e-06 05-03 16:37:39.163 6567 30025 I python : 1.9063839e-06 8.1505374e-07] 05-03 16:37:39.163 6567 30025 I python : [9.3079507e-01 9.7354960e-01 1.9993705e-01 ... 2.2026243e-06 05-03 16:37:39.163 6567 30025 I python : 2.1150458e-06 8.2314176e-07] 05-03 16:37:39.163 6567 30025 I python : [9.7550833e-01 9.7113740e-01 1.7806411e-01 ... 1.6881464e-06 05-03 16:37:39.163 6567 30025 I python : 1.1987436e-06 6.0122733e-07]] 05-03 16:37:39.164 6567 30025 I python : Scores: [4.7007226e-05 8.5895088e-05 4.0709638e-05 ... 3.7478853e-06 3.3213526e-06 05-03 16:37:39.164 6567 30025 I python : 2.8609834e-06] 05-03 16:37:39.180 6567 30025 I python : Indices after NMS: [[7601] 05-03 16:37:39.180 6567 30025 I python : [7853] 05-03 16:37:39.180 6567 30025 I python : [7668] 05-03 16:37:39.180 6567 30025 I python : [6563] 05-03 16:37:39.180 6567 30025 I python : [7878] 05-03 16:37:39.180 6567 30025 I python : [7088] 05-03 16:37:39.180 6567 30025 I python : [7262] 05-03 16:37:39.180 6567 30025 I python : [7287] 05-03 16:37:39.180 6567 30025 I python : [6536] 05-03 16:37:39.180 6567 30025 I python : [2125] 05-03 16:37:39.180 6567 30025 I python : [6607] 05-03 16:37:39.180 6567 30025 I python : [3547] 05-03 16:37:39.180 6567 30025 I python : [6808] 05-03 16:37:39.180 6567 30025 I python : [5907] 05-03 16:37:39.180 6567 30025 I python : [7688] 05-03 16:37:39.180 6567 30025 I python : [1336] 05-03 16:37:39.180 6567 30025 I python : [6896] 05-03 16:37:39.180 6567 30025 I python : [6972] 05-03 16:37:39.180 6567 30025 I python : [7068] 05-03 16:37:39.180 6567 30025 I python : [2867] 05-03 16:37:39.180 6567 30025 I python : [4295] 05-03 16:37:39.180 6567 30025 I python : [7497] 05-03 16:37:39.180 6567 30025 I python : [5086] 05-03 16:37:39.180 6567 30025 I python : [ 448] 05-03 16:37:39.180 6567 30025 I python : [4266] 05-03 16:37:39.180 6567 30025 I python : [5798] 05-03 16:37:39.180 6567 30025 I python : [ 705] 05-03 16:37:39.180 6567 30025 I python : [7077] 05-03 16:37:39.180 6567 30025 I python : [6478] 05-03 16:37:39.180 6567 30025 I python : [6483] 05-03 16:37:39.180 6567 30025 I python : [7242]] 05-03 16:37:39.181 6567 30025 I python : [INFO ] [Base ] Leaving application in progress... 05-03 16:37:39.181 6567 30025 I python : Traceback (most recent call last): 05-03 16:37:39.181 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/app/main.py", line 165, in 05-03 16:37:39.181 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/app.py", line 956, in run 05-03 16:37:39.182 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/base.py", line 574, in runTouchApp 05-03 16:37:39.182 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/base.py", line 339, in mainloop 05-03 16:37:39.182 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/base.py", line 383, in idle 05-03 16:37:39.182 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/base.py", line 334, in dispatch_input 05-03 16:37:39.182 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/base.py", line 263, in post_dispatch_input 05-03 16:37:39.183 6567 30025 I python : File "kivy/_event.pyx", line 731, in kivy._event.EventDispatcher.dispatch 05-03 16:37:39.183 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/core/window/__init__.py", line 1709, in on_motion 05-03 16:37:39.183 6567 30025 I python : File "kivy/_event.pyx", line 731, in kivy._event.EventDispatcher.dispatch 05-03 16:37:39.183 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/core/window/__init__.py", line 1726, in on_touch_down 05-03 16:37:39.184 6567 30025 I python : File "kivy/_event.pyx", line 731, in kivy._event.EventDispatcher.dispatch 05-03 16:37:39.184 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/uix/screenmanager.py", line 1210, in on_touch_down 05-03 16:37:39.184 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/uix/widget.py", line 589, in on_touch_down 05-03 16:37:39.184 6567 30025 I python : File "kivy/_event.pyx", line 731, in kivy._event.EventDispatcher.dispatch 05-03 16:37:39.184 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/uix/relativelayout.py", line 306, in on_touch_down 05-03 16:37:39.185 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/uix/widget.py", line 589, in on_touch_down 05-03 16:37:39.185 6567 30025 I python : File "kivy/_event.pyx", line 731, in kivy._event.EventDispatcher.dispatch 05-03 16:37:39.185 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/uix/widget.py", line 589, in on_touch_down 05-03 16:37:39.185 6567 30025 I python : File "kivy/_event.pyx", line 731, in kivy._event.EventDispatcher.dispatch 05-03 16:37:39.185 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/platform/build-arm64-v8a/build/python-installs/myapp/arm64-v8a/kivy/uix/behaviors/button.py", line 151, in on_touch_down 05-03 16:37:39.186 6567 30025 I python : File "kivy/_event.pyx", line 727, in kivy._event.EventDispatcher.dispatch 05-03 16:37:39.186 6567 30025 I python : File "kivy/_event.pyx", line 1307, in kivy._event.EventObservers.dispatch 05-03 16:37:39.186 6567 30025 I python : File "kivy/_event.pyx", line 1231, in kivy._event.EventObservers._dispatch 05-03 16:37:39.186 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/app/main.py", line 139, in scan_and_analyze 05-03 16:37:39.186 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/app/main_model.py", line 13, in main 05-03 16:37:39.187 6567 30025 I python : File "/home/jr/Desktop/Kivy/main/.buildozer/android/app/model_tflite.py", line 71, in get_positions 05-03 16:37:39.187 6567 30025 I python : ValueError: not enough values to unpack (expected 4, got 1)`


Подробнее здесь: https://stackoverflow.com/questions/784 ... n-kivy-app