lojzezust/wasr-t

Temporal WaSR-T model for maritime obstacle detection via semantic segmentation

wasr-t's Issues

WaSR-T License

Thank you @lojzezust for sharing your work on WaSR-T. It will be useful if a License is provided for this repository. I am looking forward to it.

Thanks in advance.

How can I use the module in an obstacle avoidance system in real life?

Hi, I'm trying to use the WaSR-T segmentation output in Jetson Nano to make an obstacle avoidance system. Is there any recommendation on how to do this? For example, I was thinking of dividing the frames into a 6x6 matrix and determining if there is a percentage greater than 20% of obstacles in it to consider it a blocked frame. And so try to make decisions, but I find some limitations such as I can not predict the trajectory of the obstacles. How could you assess the risk of collision? I have the Jetson Nano connected to a pixhawk that I put in guided mode and manage it through a script made with dronekit

Error to running inference after training the model

Hi,
i had the problem to running the inference after trying to train your models, use the train.py script following your instruction.

this the error message:
oprek@1dd801f98bc4:~/Wasr/WaSR-T$ python predict_sequential.py --sequence-dir examples/sequence --hist-len 5 --weights output/models/my_wasr_v36/model.pth --output-dir output/predictions --resize none --fp16 --gpu -1
Namespace(fp16=True, gpus='-1', hist_len=5, output_dir='output/predictions', resize=[None], sequence_dir='examples/sequence', weights='output/models/my_wasr_v36/model.pth')
Traceback (most recent call last):
File "predict_sequential.py", line 100, in
main()
File "predict_sequential.py", line 96, in main
run_inference(args)
File "predict_sequential.py", line 79, in run_inference
state_dict = load_weights(args.weights)
File "/home/oprek/Wasr/WaSR-T/wasr_t/utils.py", line 34, in load_weights
if 'model' in state_dict:
TypeError: argument of type 'WaSRT' is not iterable

Execute in Jetson Nano 4GB

Hi guys,
I had problem when execute the command
pip install -r requirements.txt
With the following output

(env) jetbot@nano-4gb-jp45:~ /WaSR-T$ pip install -r requirements.txt
Collecting torch==1.8.1 (from -r requirements.txt (line 1))
Could not find a version that satisfies the requirement torch==1.8.1 (from -r requirements.txt (line 1)) (from versions: )
No matching distribution found for torch==1.8.1 (from -r requirements.txt (line 1))
(env) jetbot@nano-4gb-jp45:~ /WaSR-T$ pip install torch
Collecting torch
Could not find a version that satisfies the requirement torch (from versions: )
No matching distribution found for torch

I have a virtual enviroment with python 3.8 install on ubuntu 18.04 for arm arch.