Project: Machine Learning models for image classification, object detection and segmentation in Virtual Reality

[DaniilOr (Daniil)]

My work for the past week:

• Made MR with detection inference feat inference: add detection inference (!2) · Merge requests · LibreHealth / LibreHealth Radiology / lh-radiology-vr-models · GitLab
• Start experimenting with classification task

Tasks for next week:

• Write inference for tools classification
• Start scene classification

[DaniilOr (Daniil)]

My work for the past week:

• Did a MR: Classification (!3) · Merge requests · LibreHealth / LibreHealth Radiology / lh-radiology-vr-models · GitLab
• Start building a model for scene classification (the data loader works too slow, I am trying to find out a way to speed it up)

Tasks for the next week:

• Finish experiments with scene classification
• Prepare inference for it
• Prepare inference for video stream (running all steps for detection and segmentation for a video) @sunbiz correct me if this part is not detailed enough
• Complete notebooks with Detection inference results

[pri2si17 (Priyanshu Sinha)]

Hi @DaniilOr I won’t be able to attend todays meeting. Please provide update on the pipeline I asked you to create with some output (maybe you can include images in reply to this thread.)

[DaniilOr (Daniil)]

My work for the past week:

• Prepared the video annotator pipeline: Files · annotator · Daniil / lh-radiology-vr-models · GitLab
• Added examples of detection: experiments · detection · Daniil / lh-radiology-vr-models · GitLab Some of the examples can be shown here, I guess. TL DR: I have experimented with different distortion parameters, the model detects the surgical tools well for almost all parameters.

Tasks for the next week:

• Do a major code refactoring (but merging my MRs is needed), write documentation

[pri2si17 (Priyanshu Sinha)]

Hi @DaniilOr , thanks for the update. Will review and merge your PR by this week.

[r0bby (Robby O'Connor)]

Your weekly report is due.

Google Summer of Code 2022 Weekly Status Reports

We are going to try something new that we introduced in 2020. This is in addition to your blog post. The goal is to catch things early if you are behind in your timeline early. @gsoc2022contributors: Answer the following questions in a once-weekly post on Fridays: What progress have you made this week? What do I plan to do next week? Have you had any blockers or issues that are impeding your project? The point of these is to catch any issues you are having and to handle them early. Any que…

[DaniilOr (Daniil)]

Done. Please sorry for a small delay, I was in a train

[DaniilOr (Daniil)]

My work for the past week:

• Local refactoring
• Optimisation of video processing pipeline

Tasks for the next week:

• Make the segmentation results easier to understand (because of grayscale the classes on the output are hard to identify)
• Prepare the detection demo (and upload here the video)
• Improve the speed of pipeline
• Write a comprehensive guide for running the pipeline and reusing the code

@pri2si17 correct me if something is missing or wrong here

[DaniilOr (Daniil)]

Here are some examples of videos: DropMeFiles – free one-click file sharing service Now there is no need to move between frames using the keyboard, processing is faster and the segmentation results are easy to read. I hope this quality is fine

[r0bby (Robby O'Connor)]

No issues – I just remind you – me needing to remind you isn’t going to go against you…Now, if I’m reminding you on Sundays, that’s an issue; otherwise, you’re fine.

[DaniilOr (Daniil)]

My work for the past week:

• Major global refactoring
• Improving the processing speed
• Adding examples for detection

Tasks for the next week:

• Change the segmentation video processor so that only the surgical tool is segmented

[DaniilOr (Daniil)]

@pri2si17

Hi! Is it the correct output?

[DaniilOr (Daniil)]

@pri2si17 can you review Final submission (!5) · Merge requests · LibreHealth / LibreHealth Radiology / lh-radiology-vr-models · GitLab and merge it, if everything is fine?

[pri2si17 (Priyanshu Sinha)]

Hi @DaniilOr, will have a look at it.

[DaniilOr (Daniil)]

@pri2si17 I also would like to start working towards the final submission. Are the any fixes that I have to add before moving to this part? Also, can you review the final submission post when it is ready?

[pri2si17 (Priyanshu Sinha)]

Hi @DaniilOr your large MR is having merge conflicts. Can you please rebase and fix it? And thanks for the good work!

[DaniilOr (Daniil)]

Thanks. Conflicts are resolved.

[pri2si17 (Priyanshu Sinha)]

Hi @DaniilOr , thanks for quickly resolving the conflicts. I wont be able to meet, but I want you to make an architecture diagram of the whole system and update the README with instructions to run.

[DaniilOr (Daniil)]

Ok, I will do it.

Thanks for this recommendation

[DaniilOr (Daniil)]

I have added something like a system diagram here: Summing up my GSoC experience I will also include it into Readme. What else should I add?