PRATIK MAZUMDER

Results 15 comments of PRATIK MAZUMDER

FPAIT - How are fully connected layers generating 3D weights for Convolutional layer outputs?

@D-X-Y thanks for pointing that out. Also when will you be releasing the training code?

FPAIT - How are fully connected layers generating 3D weights for Convolutional layer outputs?

@D-X-Y can you please provide a skeleton for fetching the data. I am getting errors trying to do so. I tried the following `from lib.datasets.qatask import MetaQATask` `mqat = MetaQATask(dataset='toronto-coco-qa',pth_data='./data/Toronto-COCO-QA/object.pth',num_classes=5,num_samples_per_class=5)`...

FPAIT - How are fully connected layers generating 3D weights for Convolutional layer outputs?

Thanks a lot

Code Request of FPAIT

Hi, could you please release the code for training and testing of FPAIT.

test_semisup gives Validation error rate around 0.12 for svhn after training for 100 epochs

Ok so is it normal that I am getting 12% on the same code?

test_semisup gives Validation error rate around 0.12 for svhn after training for 100 epochs

No I have not made any changes. Infact I am getting around 15% for the netE_epoch_90.pth provided by default in this repository using the test_semisup.py provided in the repository. I...

test_semisup gives Validation error rate around 0.12 for svhn after training for 100 epochs

Can you please check the validation error for my saved model.

test_semisup gives Validation error rate around 0.12 for svhn after training for 100 epochs

ok thanks

test_semisup gives Validation error rate around 0.12 for svhn after training for 100 epochs

Hi, did u check the models?

Confusion Regarding belga2toplogo result

This is the entry in **best_iou_val.txt** for **belga2toplogo** Best score epoch: 420 acc cls: 0.6022 acc all: 0.6072 rank sample mean: 0.8964 rank cls mean: 0.8940 acc tr cls: 0.8399...