pytorch-semseg
pytorch-semseg copied to clipboard
meetps
Semantic Segmentation Architectures Implemented in PyTorch
In train.py, the output starts from "Iter [2/1000]...". I went into the code, and I found out that i starts from 0 (assuming no "resume training" scheme). In the first...
Hi, guys: Nice job！ I tried this project with default settings, and it works well except that the logging module would print messages both into logfile and console like the...
I use this code to train segnet on camvid, but the results can't reach the paper performance. I found that you didn't use median-frequency balancing to training. Moreover, unlabelled class...
I have a question about line 67 in test.py: model = get_model(model_name, n_classes, version=args.dataset) model_name is a string but the first argument in get_model is a dict so the code...
Hello, When I run test.py, I get the following error: ``` Traceback (most recent call last): File "/home/ubuntu/pytorch-semseg/test.py", line 170, in test(args) File "/home/ubuntu/pytorch-semseg/test.py", line 67, in test model =...
Hi, I previous opened an issue about ICNet evaluation mean-iou, it seems stopped at 0.25 approximately. ``` 5it [00:01, 2.34it/s]WARN: resizing labels yielded fewer classes 250it [00:23, 10.78it/s] Overall Acc:...
Has anyone implemented this? If not, I can write the data loader for this and create a PR. Thanks!
I added this code snippet to the UNet model so that I can use bilinear interpolation: ``` if __name__=="__main__": x = torch.randn([5, 3, 300, 300]) model = unet(is_deconv=False) y =...
I want to train ade20k on pspnet i use your train_cnn, but it appears like this one File "/home/victor/catkin_ws/src/semantic_slam/semantic_cloud/include/ptsemseg/loss.py", line 74, in multi_scale_cross_entropy2d scale_weight = torch.pow(scale * torch.ones(n_inp), torch.arange(n_inp)) RuntimeError:...
ENet is a fast neural network for semantic segmentation, I use the code from [other's](https://github.com/davidtvs/PyTorch-ENet/blob/master/models/enet.py), and I simply train the module in CamVid Dataset, it seems working well, can it...
