potterhsu / easy-faster-rcnn.pytorch

An easy implementation of Faster R-CNN (https://arxiv.org/pdf/1506.01497.pdf) in PyTorch.
MIT License
165 stars 57 forks source link

Checkpoint for Resnet50 and Resent18? #13

Open piyushy1 opened 5 years ago

piyushy1 commented 5 years ago

By any chance, do you have the checkpoints for resnet50 or resnet18? I'm using the inference part of the script because my GPU isn't strong enough for training it. It would be really helpful if you have it.

Thanks!

potterhsu commented 5 years ago

Here are more trained models: Resnet-18 trained on VOC2007, got mAP = 0.6791 Resnet-50 trained on VOC2007, got mAP = 0.7438 Resnet-50 trained on COCO2017, got AP = 0.3252 Resnet-50 trained on COCO2017 with a longer schedule, got AP = 0.3447

sachingoyal0104 commented 4 years ago

Hi, Can u provide checkpoints for coco2017-person model.. Thanks

Hrqingqing commented 2 years ago

Here are more trained models: Resnet-18 trained on VOC2007, got mAP = 0.6791 Resnet-50 trained on VOC2007, got mAP = 0.7438 Resnet-50 trained on COCO2017, got AP = 0.3252 Resnet-50 trained on COCO2017 with a longer schedule, got AP = 0.3447

Thank you for your contributions. I would like to ask about the training hyperparameters for Resnet50 on VOC2007.

Hrqingqing commented 2 years ago

Here are more trained models: Resnet-18 trained on VOC2007, got mAP = 0.6791 Resnet-50 trained on VOC2007, got mAP = 0.7438 Resnet-50 trained on COCO2017, got AP = 0.3252 Resnet-50 trained on COCO2017 with a longer schedule, got AP = 0.3447

Thank you for your contributions. I would like to ask about the training hyperparameters for Resnet50 on VOC2007.

What I mean is that without loading the pre-trained model parameters.