weird behaviour in stanford car

Alibaba-MIIL / ML_Decoder

Official PyTorch implementation of "ML-Decoder: Scalable and Versatile Classification Head" (2021)

MIT License

315 stars 52 forks source link

weird behaviour in stanford car #67

Open ShihaoShao-GH opened 2 years ago

ShihaoShao-GH commented 2 years ago

Hi, thanks for your amazing work!

Here I encountered something really weird. I downloaded the tresnet_l_stanford_card_96.41.pth and tried to validate the result in stanford car datasets. It reached 99.69 for validation part and 96.80 for train part, respectively. Most likely the training and validation part were treated inversely.

Can you please check if the order is correct?

Sondosmohamed1 commented 9 months ago

Hi, I have a question: Did you reload the pretrained model, and did it work? Because, for me, I encountered an error when inferring with 'tresnet_l_stanford_card_96.41.pth'. However, when I run 'tresnet_m_COCO_224_84_2.pth', it works.