search

axinc-ai / ailia-models

The collection of pre-trained, state-of-the-art AI models for ailia SDK
2.04k stars 325 forks source link

add --seed option for aclip_clip #1189

Closed kzmogi closed 1 year ago

kzmogi commented 1 year ago

add --seed option to keep consistency of results.

without --seed option, prob value is randomly varying.

D:\ailia\ailia-models\action_recognition\action_clip>python action_clip.py -e 0
 INFO arg_utils.py (13) : Start!
 INFO arg_utils.py (163) : env_id: 0
 INFO arg_utils.py (166) : CPU
 INFO action_clip.py (273) : No text provided, using ['applauding', 'dancing', 'driving', 'driving car', 'driving truck', 'eating', 'punching', 'reading', 'surfing', 'talking phone']...
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO action_clip.py (241) : Start inference...
 INFO action_clip.py (209) : ==============================================================
 INFO action_clip.py (210) : class_count = 10
 INFO action_clip.py (213) : + idx = 0
 INFO action_clip.py (214) :   category = 4 [driving truck]
 INFO action_clip.py (215) :   prob = 0.5996775031089783
 INFO action_clip.py (213) : + idx = 1
 INFO action_clip.py (214) :   category = 3 [driving car]
 INFO action_clip.py (215) :   prob = 0.19072364270687103
 INFO action_clip.py (213) : + idx = 2
 INFO action_clip.py (214) :   category = 2 [driving]
 INFO action_clip.py (215) :   prob = 0.1810181438922882
 INFO action_clip.py (213) : + idx = 3
 INFO action_clip.py (214) :   category = 9 [talking phone]
 INFO action_clip.py (215) :   prob = 0.027360258623957634
 INFO action_clip.py (213) : + idx = 4
 INFO action_clip.py (214) :   category = 7 [reading]
 INFO action_clip.py (215) :   prob = 0.0010408649686723948
 INFO action_clip.py (216) :
 INFO action_clip.py (255) : Script finished successfully.

D:\ailia\ailia-models\action_recognition\action_clip>python action_clip.py -e 0
 INFO arg_utils.py (13) : Start!
 INFO arg_utils.py (163) : env_id: 0
 INFO arg_utils.py (166) : CPU
 INFO action_clip.py (273) : No text provided, using ['applauding', 'dancing', 'driving', 'driving car', 'driving truck', 'eating', 'punching', 'reading', 'surfing', 'talking phone']...
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO action_clip.py (241) : Start inference...
 INFO action_clip.py (209) : ==============================================================
 INFO action_clip.py (210) : class_count = 10
 INFO action_clip.py (213) : + idx = 0
 INFO action_clip.py (214) :   category = 4 [driving truck]
 INFO action_clip.py (215) :   prob = 0.622225284576416
 INFO action_clip.py (213) : + idx = 1
 INFO action_clip.py (214) :   category = 2 [driving]
 INFO action_clip.py (215) :   prob = 0.1704442948102951
 INFO action_clip.py (213) : + idx = 2
 INFO action_clip.py (214) :   category = 3 [driving car]
 INFO action_clip.py (215) :   prob = 0.16028760373592377
 INFO action_clip.py (213) : + idx = 3
 INFO action_clip.py (214) :   category = 9 [talking phone]
 INFO action_clip.py (215) :   prob = 0.04551739618182182
 INFO action_clip.py (213) : + idx = 4
 INFO action_clip.py (214) :   category = 7 [reading]
 INFO action_clip.py (215) :   prob = 0.0012974952114745975
 INFO action_clip.py (216) :
 INFO action_clip.py (255) : Script finished successfully.

with --seed option, prob keeps same value.

D:\ailia\ailia-models\action_recognition\action_clip>python action_clip.py -e 0 --seed 20230726
 INFO arg_utils.py (13) : Start!
 INFO arg_utils.py (163) : env_id: 0
 INFO arg_utils.py (166) : CPU
 INFO action_clip.py (273) : No text provided, using ['applauding', 'dancing', 'driving', 'driving car', 'driving truck', 'eating', 'punching', 'reading', 'surfing', 'talking phone']...
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO action_clip.py (241) : Start inference...
 INFO action_clip.py (209) : ==============================================================
 INFO action_clip.py (210) : class_count = 10
 INFO action_clip.py (213) : + idx = 0
 INFO action_clip.py (214) :   category = 4 [driving truck]
 INFO action_clip.py (215) :   prob = 0.6156361699104309
 INFO action_clip.py (213) : + idx = 1
 INFO action_clip.py (214) :   category = 2 [driving]
 INFO action_clip.py (215) :   prob = 0.16875791549682617
 INFO action_clip.py (213) : + idx = 2
 INFO action_clip.py (214) :   category = 3 [driving car]
 INFO action_clip.py (215) :   prob = 0.15870651602745056
 INFO action_clip.py (213) : + idx = 3
 INFO action_clip.py (214) :   category = 9 [talking phone]
 INFO action_clip.py (215) :   prob = 0.05544034019112587
 INFO action_clip.py (213) : + idx = 4
 INFO action_clip.py (214) :   category = 7 [reading]
 INFO action_clip.py (215) :   prob = 0.0011933924397453666
 INFO action_clip.py (216) :
 INFO action_clip.py (255) : Script finished successfully.

D:\ailia\ailia-models\action_recognition\action_clip>python action_clip.py -e 0 --seed 20230726
 INFO arg_utils.py (13) : Start!
 INFO arg_utils.py (163) : env_id: 0
 INFO arg_utils.py (166) : CPU
 INFO action_clip.py (273) : No text provided, using ['applauding', 'dancing', 'driving', 'driving car', 'driving truck', 'eating', 'punching', 'reading', 'surfing', 'talking phone']...
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO model_utils.py (84) : ONNX file and Prototxt file are prepared!
 INFO action_clip.py (241) : Start inference...
 INFO action_clip.py (209) : ==============================================================
 INFO action_clip.py (210) : class_count = 10
 INFO action_clip.py (213) : + idx = 0
 INFO action_clip.py (214) :   category = 4 [driving truck]
 INFO action_clip.py (215) :   prob = 0.6156361699104309
 INFO action_clip.py (213) : + idx = 1
 INFO action_clip.py (214) :   category = 2 [driving]
 INFO action_clip.py (215) :   prob = 0.16875791549682617
 INFO action_clip.py (213) : + idx = 2
 INFO action_clip.py (214) :   category = 3 [driving car]
 INFO action_clip.py (215) :   prob = 0.15870651602745056
 INFO action_clip.py (213) : + idx = 3
 INFO action_clip.py (214) :   category = 9 [talking phone]
 INFO action_clip.py (215) :   prob = 0.05544034019112587
 INFO action_clip.py (213) : + idx = 4
 INFO action_clip.py (214) :   category = 7 [reading]
 INFO action_clip.py (215) :   prob = 0.0011933924397453666
 INFO action_clip.py (216) :
 INFO action_clip.py (255) : Script finished successfully.