michael-xiu / icdar-sroie Goto Github PK

Can the code run in win10 + pytorch 1.1? Thank you

Hi Michael,
Thanks for sharing your code. We are trying to use your code and train on the SROIE 2019 receipt dataset. However, we are getting very poor results. Can you assist us with providing your last checkpoint?
Thanks
Ahmed

This is the current train and val loss:

epoch:184 train loss:3.999357543253462 val loss:4.058299102783203
epoch:185 train loss:4.015532674487934 val loss:4.1186825561523435
epoch:186 train loss:4.000330889681215 val loss:4.067411651611328
epoch:187 train loss:4.003223073264326 val loss:4.099950981140137
epoch:180 train loss:4.001502086239528 val loss:4.044377269744873
epoch:181 train loss:3.9814491708346096 val loss:4.074534072875976
epoch:182 train loss:4.092637693624131 val loss:4.092916450500488
epoch:183 train loss:3.971474315878159 val loss:4.15595552444458
epoch:184 train loss:4.002527025892413 val loss:4.103062620162964
epoch:185 train loss:3.9752465238587034 val loss:4.050528306961059
epoch:186 train loss:3.962182341716849 val loss:4.080765295028686
epoch:187 train loss:4.000311766606996 val loss:4.108867740631103
epoch:188 train loss:3.9928070542816316 val loss:4.080308256149292
epoch:189 train loss:3.9871446995092508 val loss:4.109854030609131
epoch:190 train loss:3.9957085393629534 val loss:4.157553844451904
epoch:191 train loss:3.943965378696232 val loss:4.033948841094971
epoch:192 train loss:3.9847304602827687 val loss:4.108264808654785
epoch:192 train loss:3.9862627324566073 val loss:4.069806537628174

RuntimeError: Expected object of scalar type Byte but got scalar type Bool for argument #2 'other'
ICDAR/ICDAR-SROIE/task1_SSD/model.py", line 536, in detect_objects

Traceback (most recent call last):
File "eval.py", line 141, in
evaluate(test_loader, model, test_img_num)
File "eval.py", line 87, in evaluate
top_k=200, original_image = images, max_OCR_overlap=0.2, max_OCR_ratio = 0.2)
File "/ctpn/sroie/sroie/task1/model.py", line 588, in detect_objects
[original_image.width, original_image.height, original_image.width,
AttributeError: 'Tensor' object has no attribute 'width'