search

CAMMA-public / rendezvous

A transformer-inspired neural network for surgical action triplet recognition from laparoscopic videos.
Other
24 stars 8 forks source link

reproduced result #3

Closed gshuangchun closed 2 years ago

gshuangchun commented 2 years ago

When I used the default hyperparameters, I got a bad result by using pytorch. Could you please release the hyperparameters used by the best results?

My results are as follows: python run.py --train --kfold 1 --gpu 0

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8028 | 0.5914 | 0.3454 | 0.3218 | 0.2888 | 0.2299

python run.py --train --kfold 2 --gpu 0

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8129 | 0.5572 | 0.3334 | 0.2865 | 0.2568 | 0.2033

python run.py --train --kfold 3 --gpu 0

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8561 | 0.5684 | 0.3352 | 0.3892 | 0.2961 | 0.2475

python run.py --train --kfold 4 --gpu 0

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8188 | 0.5247 | 0.3109 | 0.2975 | 0.2667 | 0.2086

nwoyecid commented 2 years ago

Please, check the published papers for their used hyper-parameters. Also, note, the original papers use different augmentation style per epoch, which is different from the simplified dataloader in the git repo.

csEylLee commented 1 year ago

When I used the default hyperparameters, I got a bad result by using pytorch. Could you please release the hyperparameters used by the best results?

My results are as follows:

python run.py --train --kfold 1 --gpu 0 Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8028 | 0.5914 | 0.3454 | 0.3218 | 0.2888 | 0.2299

python run.py --train --kfold 2 --gpu 0

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8129 | 0.5572 | 0.3334 | 0.2865 | 0.2568 | 0.2033

python run.py --train --kfold 3 --gpu 0

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8561 | 0.5684 | 0.3352 | 0.3892 | 0.2961 | 0.2475

python run.py --train --kfold 4 --gpu 0

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8188 | 0.5247 | 0.3109 | 0.2975 | 0.2667 | 0.2086

Have you solved it? I meet the same problem. 前辈，请问您这个问题解决了吗？我遇到了相同的问题。

DoctorMango commented 7 months ago

The same question, have you solved it, I tried some parameters but I don't get the same results. Could you please share your hyperparameters? Especially the pretrain_dir,initial_learning_rates, I can't make sure my parameters.

DoctorMango commented 7 months ago

When I used the default hyperparameters, I got a bad result by using pytorch. Could you please release the hyperparameters used by the best results?

My results are as follows:

python run.py --train --kfold 1 --gpu 0 Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8028 | 0.5914 | 0.3454 | 0.3218 | 0.2888 | 0.2299

python run.py --train --kfold 2 --gpu 0

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8129 | 0.5572 | 0.3334 | 0.2865 | 0.2568 | 0.2033

python run.py --train --kfold 3 --gpu 0

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8561 | 0.5684 | 0.3352 | 0.3892 | 0.2961 | 0.2475

python run.py --train --kfold 4 --gpu 0

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8188 | 0.5247 | 0.3109 | 0.2975 | 0.2667 | 0.2086

hello, can i get your hyperparameters, I can't reproduce the paper result, so I want to reproduce your result first.