Inputs required for training

Hi @wvangansbeke ,

Great work! I downloaded the sample Tusimple dataset provided in the repo. It has only two folders namely images and groundtruth. After this, I ran the code using below command python main.py --image_dir /path/to/image/folder --gt_dir /path/to/ground_truth/folder --end_to_end True
and training process did not start and got the below error: (ld_env) $ python main.py --image_dir /home/teai/DVA/LaneDetection_End2End/DATASET/images/ --gt_dir /home/teai/DVA/LaneDetection_End2End/DATASET/ground_truth/ --end_to_end False size train loader is 2901 size valid loader is 725 Flipping images randomly: False End to end lane detection: False size test loader: 2782 Init weights in network with [kaiming]

Args:Namespace(activation_layer='square', activation_net='relu', batch_size=8, channels_in=3, clas=False, clip_grad_norm=0, crop_size=80, cudnn=True, dataset='lane_detection', draw_testset=False, end_to_end=False, evaluate=False, flip_on=False, gamma=0.0, gt_dir='/home/teai/DVA/LaneDetection_End2End/DATASET/ground_truth/', image_dir='/home/teai/DVA/LaneDetection_End2End/DATASET/images/', json_file='Labels/Curve_parameters.json', layers=18, learning_rate=0.0001, list=[954, 2789], loss_policy='area', lr_decay=False, lr_decay_iters=30, lr_policy=None, mask_percentage=0.3, mod='erfnet', nclasses=2, nepochs=500, niter=50, niter_decay=400, no_cuda=False, no_dropout=False, no_mapping=False, no_ortho=False, no_tb=True, norm='batch', num_train=3626, nworkers=8, optimizer='adam', order=2, pool=True, pretrain_epochs=20, pretrained=False, print_freq=500, reg_ls=0, resize=256, resume='', save_freq=100, save_path='Saved/Mod_erfnet_opt_adam_loss_area_lr_0.0001_batch_8_end2end_False_chol_False_lanes_2_pretrainFalse_clasFalse_mask0.3_flip_onFalse_activation_square', skip_epochs=10, split_percentage=0.2, start_epoch=0, test_dir='/usr/data/tmp/Lane_Detection/TESTSET/', test_mode=False, use_cholesky=False, val_batch_size=8, weight_class=1, weight_decay=0, weight_fit=1, weight_funct='none', weight_init='kaiming', weight_seg=30)

Init model: 'erfnet' Number of parameters in model ERFNET is 2.064M

=> Start train set for EPOCH 1 Saving to: Saved/Mod_erfnet_opt_adam_loss_area_lr_0.0001_batch_8_end2end_False_chol_False_lanes_2_pretrainFalse_clasFalse_mask0.3_flip_onFalse_activation_square 0it [00:00, ?it/s]Traceback (most recent call last): File "main.py", line 552, in main() File "main.py", line 243, in main for i, (input, gt, lanes, idx, gt_line, gt_horizon, valid_points) in tqdm(enumerate(train_loader)): File "/home/teai/DVA/ld_env/lib/python3.6/site-packages/tqdm/std.py", line 1102, in iter for obj in iterable: File "/home/teai/DVA/ld_env/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 819, in next return self._process_data(data) File "/home/teai/DVA/ld_env/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 846, in _process_data data.reraise() File "/home/teai/DVA/ld_env/lib/python3.6/site-packages/torch/_utils.py", line 369, in reraise raise self.exc_type(msg) TypeError: Caught TypeError in DataLoader worker process 0. Original Traceback (most recent call last): File "/home/teai/DVA/ld_env/lib/python3.6/site-packages/torch/utils/data/_utils/worker.py", line 178, in _worker_loop data = fetcher.fetch(index) File "/home/teai/DVA/ld_env/lib/python3.6/site-packages/torch/utils/data/_utils/fetch.py", line 47, in fetch return self.collate_fn(data) File "/home/teai/DVA/ld_env/lib/python3.6/site-packages/torch/utils/data/_utils/collate.py", line 82, in default_collate raise TypeError(default_collate_err_msg_format.format(elem_type)) TypeError: default_collate: batch must contain tensors, numpy arrays, numbers, dicts or lists; found <class 'NoneType'> 0it [00:00, ?it/s]

Do you have any idea on how to recover from this error ? Also do I need entire Tusimple dataset to run this or subset which was specified in repo is enough?

Thanks in advance..

Regards Abhi

wvangansbeke / LaneDetection_End2End

Inputs required for training #59