arise-lab / deeptest Goto Github PK

Hello and thank you for the deepTest code,

Unfortunately, I couldn't find the HMB_3 bag file so I wanted to use the other HMB_1, HMB2, files from CH2_002 dataset, however I get "list index out of bounds" errors when I try to do that. How can I use another dataset with deepTest?

Kind regards,

I tried running this command
python generate_hmb3.py --bagfile HMB_3.bag

but there was no HMB3.bag file so I tried to retrieve it from the link provided in the README (https://github.com/udacity/self-driving-car/blob/master/datasets/CH2/HMB_3.bag.tar.gz.torrent) but I only downloaded a torrent file. I tried using a bitTorrent client to unpack it but it does not seem to be seeding.

I would be really appreciate your help if you could point me in the right direction of getting this HMB_3.bag file so that I could carry out the DeepTest algorithm. Thank you for also providing this source code for the community to learn.

Test dataset not downloading. Is there an alternative download method?

Hi @yuchi1989,
Greetings! Thanks for sharing the source code. I am trying to execute the code for Guided Transformation (Algorithm # 1 from the deepTest paper). I noticed that the rambo_guided.py uses a set of image transformation values (line # 358 - line #354) that is different from the values discussed in the paper (Table 4). Can you please help me understand why a different set of transformation values was used for Guided Transformation experiments?

Thanks,
Jagan

Hi @yuchi1989,
I follow the readme to prepare the datasets. However, I cannot download the HMB_3 experiment since there is no seed for the torrent link (https://github.com/udacity/self-driving-car/blob/master/datasets/CH2/HMB_3.bag.tar.gz.torrent).

Do you know any other ways to access the data? Thank you for your consideration.

I use the rambo model(rambo_reproduce.py) to test the CH2_final dataset, I test by your description in readme.md. Download the model weight and train_mean file in https://github.com/udacity/self-driving-car/tree/master/steering-models/evaluation But i get mse 0.30 rather 0.06. Do you have some thoughts about this problem?