xhwdlgy5201 / superthermal-thermal-matching Goto Github PK

Hello,

Thank you for sharing your implementation. I am having some trouble running python batch_draw_matching_points.py --resume KAIST.pth.tar --folder test_input

The resulting error is as followed while using pytorch 2.01:

Loading model from KAIST.pth.tar ['test_input/II01985.jpg', 'test_input/II01980.jpg'] Image 1 path: test_input/II01980.jpg Image 2 path: test_input/II01985.jpg /home/jonathansel/SuperThermal-thermal-matching/model/thermal_det_module.py:197: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than tensor.new_tensor(sourceTensor). psf = im1w_score.new_tensor( Traceback (most recent call last): File "batch_draw_matching_points.py", line 95, in <module> kp1, des1, img1 = model.detectAndCompute(img1_path, device, (240, 320)) File "/home/jonathansel/SuperThermal-thermal-matching/model/thermal_net_so.py", line 205, in detectAndCompute _, kp, des = self.inference(img, img_info, img_raw) File "/home/jonathansel/SuperThermal-thermal-matching/model/thermal_net_so.py", line 154, in inference im_score = self.det.process(im_rawsc)[0] File "/home/jonathansel/SuperThermal-thermal-matching/model/thermal_det_module.py", line 202, in process im1w_score = F.conv2d( RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same

If you have any suggestions on how to fix please let me know, thank you!