Details about data preprocessing.

[linjing7]

Hi, thanks for your excellent work.

In addition to the feature map, I want to take the RGB image as input of the network. However, the provided preprocessed data do not include the RGB image. I wonder if you could provide a preprocessed data that contain the RGB image.

Or could you please tell me how to preprocess the data? You mention in 'data.md' that "You may need to change details (ex. scale), so check comments in {dataset_name}_utils.py files.". So I worry about that I could not correctly preprocess the data if there is something wrong with the details.

Thank you very much!

[hongsukchoi]

Hi @linjing7

Are you trying to build a new network on top of TCMR?

[linjing7]

Yes, I'm trying. If it works, I'll cite TCMR and thank you in the acknowledgement :)

[hongsukchoi]

Hi @linjing7,

OK, I will tell you how to preprocess data. I asked you because there's one thing you should know about the InstaVariety. It is preprocessed by VIBE's authors and the images are saved in tf-record. I don't know the details about extracting images from them but you could do it, though it will take some time.

Anyway, to preprocess other datasets by yourself, note two things: 'scale' and 'occ'

• 'scaleX' is added to my preprocessed file's name, when the scale is different from 1.3. https://github.com/hongsukchoi/TCMR_RELEASE/blob/ade5ab5e2bd8b1a26338406c0b272d2a16f85070/lib/data_utils/mpii3d_utils.py#L327 According to the 'X' value, change the scale parameter of extract_features in each {dataset_name}_utils.py, .

• If 'occ' is added to the file name, change the occ parameter of get_single_image_crop here. https://github.com/hongsukchoi/TCMR_RELEASE/blob/ade5ab5e2bd8b1a26338406c0b272d2a16f85070/lib/data_utils/_feature_extractor.py#L79

[linjing7]

Okay, thank you very much.