infer pics with different size

Ugness / PiCANet-Implementation

Pytorch Implementation of PiCANet: Learning Pixel-wise Contextual Attention for Saliency Detection

MIT License

178 stars 40 forks source link

infer pics with different size #19

Closed XChuanLee closed 4 years ago

XChuanLee commented 5 years ago

truely thank you for this implement ^=^ , but it confused me when i want to infer pics with different size. I run this commend: 'python image_test.py --model_dir=./model/36epo_383000step.ckpt --dataset=./../test/raw_image/ --save_dir=./../test/result_PiCA --batch_size=1' and I delete 'transforms.Resize((224, 224))' in CustomDataset but there is always an error. how can i infer pics with their original size?

Ugness commented 5 years ago

I suggest you to inference with size 224 x 224 and resize the output (saliency map) to fit with the original size of the input image.

zzzzhe commented 5 years ago

@Ugness I want a solution that outputs the original hd image. Any Suggestions?