Training from scratch

[saurabbhsp]

Hi, I want to train the model from scratch for NVGesture dataset and EgoHands dataset for depth modality. I tried with modality as depth but it throws an error (Index out of range), while for RGB the error remains constant.

PS:- This is the command used for training.

python main.py --root_path hand-gesture-recognization/Real-time-GesRec
--video_path dataset/Nvidia
--annotation_path annotation_nvGesture/nvall_but_None.json
--result_path results/nvGesture_RGB_all
--dataset nvgesture
--n_classes 25
--model resnext
--model_depth 101
--resnet_shortcut B
--version 1.1
--width_mult 1.0
--train_crop random
--learning_rate 0.01
--sample_duration 16
--modality RGB
--batch_size 64
--n_threads 16
--checkpoint 1
--n_val_samples 30
--n_epochs 3000
--test_subset test

Thanks