Add bts_sequence for pytorch

[kHarshit]

• Added bts_sequence.py for testing multiple images in a folder.

Note: Currently it's not using DataLoader.

---

Working result:

(pylatest) [kharshit@temp28 pytorch]$ CUDA_VISIBLE_DEVICES=0 python bts_sequence.py --image_path /nfs/interns/kharshit/DepthEstimation/bts/dataset/test_images/
Testing 17 files with ../models/bts_nyu_v2_pytorch_densenet161/model
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 17/17 [00:01<00:00,  9.73it/s]
Elapesed time: 1.7538671493530273
Done.
Saving result pngs..
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 17/17 [00:00<00:00, 29.96it/s]

[rsj007]

Hello, I am running your bts_sequence.py with eigen model on an image captured by a camera. When I run it, it shows error as below. I don't know where to import this. Thanks a lot!!

[kHarshit]

@rsj007 you need to provide the model path with command line argument --model_name <model path>

[rsj007]

@rsj007 you need to provide the model path with command line argument --model_name <model path>

Thanks. I solved it.
Can I test an image with size 1920*1200?

[kHarshit]

As far as I remember, your image size should be a multiple of 32. Based on the GPU RAM, you can try testing it.

[rsj007]

As far as I remember, your image size should be a multiple of 32. Based on the GPU RAM, you can try testing it.

Ok! I noticed that the height of an image will be divised by 16,8,4. Maybe a multiple if 16?