-
Notifications
You must be signed in to change notification settings - Fork 9
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Result of 'Ours TC Visible' in Table 2 #2
Comments
Hi, I have the same question with u.But the cfg/yolov3_kaist.cfg doesn't have the 'condconv' module, so I am very confused about it.I guess 'Ours TC visible' was producted by TC res groups/TC all + ‘kaist_visible_detector.weigths'. |
Hi, Ours TC Visible (task-conditioned visible) used cfg/yolov3_kaist_tc_det.cfg. Any methods that have 'tc" name will use the cfg file with tc. Thanks! |
Hi,I have a question about the difference between 'Ours TC visible' and 'Ours TC Det'. Could you please help to clarify,thanks! |
Hi, Thank you very much for your reply. So you mean 'Ours TC Visible' is produced using cfg/yolov3_kaist_tc_det.cfg and 'kaist_visible_detector.weigths'? Did you train this model using only RGB images from KAIST? Thanks a lot! |
Maybe the weights are different. I think they are trained using different traning sets. |
Hi, When I tried to run with 'cfg/yolov3_kaist_tc_det.cfg' and 'kaist_visible_detector.weigts', I got the following error: But when I used 'cfg/yolov3_kaist.cfg' and 'kaist_visible_detector.weigts', it worked and gave reasonable results. However, you said you used 'cfg/yolov3_kaist_tc_det.cfg' . Have you encountered any problems? |
Hi, Ours TC visible is to train on visible-only images of KAIST dataset with task-conditioned network (starting from yolov3.weight), while Ours TC Det is to train on thermal images of KAIST dataset with task-conditioined network (starting from visible weight from previous one). |
Hi the problem because of the wrong return parameter between models (normal model and task-conditioned (TC) model). You can not load the normal weight (without tc) to the tc.cfg, you should load the tc.weight to corresponding tc.cfg because with the TC model, the output return of the network is different from normal model, so we can not use output[0]['x'].data.size(0). The easiest way to mitigate this problem is that you should load the corresponding *.cfg and *.weight. Good luck! |
Hello, thank you for your reply. I had the same problem when I trained with the FLIR dataset. I thought that cfg and model should correspond, did you have the same problem, or could you give me some advice? Thank you! |
When I download the source code directly, use cfg / yolov3_ kaist_ tc_ det.cfg corresponding with weights/yolov3_ tc_ det_ thermal. model can run in the Evaluation.py, but with weights/yolov3_ tc_ det_ thermal.weights cannot be run. I am very confused about this error! |
Hi,
Thank you very much for this very interesting work!
I have a question about the production of 'Ours TC Visible' shown in Table 2. Are you using 'cfg/yolov3_kaist.cfg' and 'kaist_sisible_detector.weights' to produce the results? Could you please help to clarify?
Bests,
Xingchen
The text was updated successfully, but these errors were encountered: