zjcv / rotnet Goto Github PK

Hi Zjcv, i sent you a couple week ago email to [email protected],
let me know if you read it cause i didnt receve any reply

model = build_recognizer(cfg, device)

How to create dataset to train my model? Thanks

Iam getting an error calling tools/train.py
" No module named 'rotnet'"
How to include?

I want to run the demo,but it reminds me that no module named rotnet,and i cant install rotnet.

ModuleNotFoundError: No module named 'rotnet'

size mismatch for backbone.first_stem.0.weight: copying a param with shape torch.Size([16, 1, 3, 3]) from checkpoint, the shape in current model is torch.Size([16, 3, 3, 3]).

Does this algorithm use a convolutional neural network like the reference?

Hi, good work! From my own testing of the model, the results are not good as it can't rotate pictures of people into vertical. Is that normal? Thanks!

Hi

You mentioned The corresponding implementation can't meet the current performance requirements, so I implement one myself in the Background section of your readme file.

Can you give more details?
Do other implementations present performance issues? accuracy issues?
In what way your implementation performs differently than d4nst/RotNet for example.

Thank you for your contribution!

I used cifar10 for pre-training, but got an error when fine-tuning with my own dataset


mbv3_small_se_hsigmoid_fmnist_224_e100.txt

ERROR: Could not find a version that satisfies the requirement zcls==0.3.2
ERROR: No matching distribution found for zcls==0.3.2

Please help!