Object detection using fastai v2

[vahuja4 (Vishal)]

Hi All,

Can someone please point me to any resources regarding this?

Sincerely,
Vishal

[mrfabulous1]

Hi vahuja4 hope your having a fantastic day!

Obect detection in fastai2 is covered in the course below.

A walk with fastai2 - Vision - Study Group and Online Lectures Megathread

Hi everyone! Each semester I lecture at the University of West Florida through one of our clubs on fastai. Normally I keep it offline but this year I have decided to live stream it along with have a dedicated megathread to discussing the course, asking questions, and networking together! This semester’s version will be using the version 2 of the fastai library (thank you to @Jeremy and @sgugger for all the amazing work you have put in!). This course is designed to be intro friendly (it’s geared …

There is a notebook you can look at as well!

Hope this helps.

Cheers mrfabulous1

[muellerzr (Zachary Mueller)]

@mrfabulous1

There is a notebook you can look at as well!

Not quite I’ll update that notebook here in the next few weeks

[mrfabulous1]

HI muellerzr
Oops! apologies just remembered fastai2 changes daily with braking changes.!
mrfabulous1

[muellerzr (Zachary Mueller)]

@vahuja4 give me a day or two and I’ll get a working example.

[gianferrarif (Francesco Gianferrari Pini)]

Hi!
I successfully ran your notebook for detection up to training, but I am struggling to get predictions on new images. Do you have a working example? Thanks!

[muellerzr (Zachary Mueller)]

I’m still working on inference for that notebook. Will be done by next Wednesday

[gianferrarif (Francesco Gianferrari Pini)]

Thanks a lot!

[j.laute (Johannes Laute)]

Good news, I played with the notebook a bit but couldn’t get show_results and mAP metrics working (I only spend an hour or so).

Let me know if I can help at all, I want to try the dilated xresnet I’m working on with object detection (in initial experiments dilation doesn’t improve segmentation with dynamic unet)

[muellerzr (Zachary Mueller)]

@j.laute

Good news, I played with the notebook a bit but couldn’t get show_results and mAP metrics working (I only spend an hour or so).

If you want to put a PR at the end showing it it’d be much appreciated!

[leaf (jakub langr)]

Hi @muellerzr thanks so much for the notebook. Just a quick question: why didn’t you use the get_annotations method? seems well suited for this and I currently struggle to get it working with the rest of your notebook. Is that why?

[muellerzr (Zachary Mueller)]

@leaf

why didn’t you use the get_annotations method?

I did, you can see we build on this once we go into img2bbox

[leaf (jakub langr)]

my bad, got it now. it looked like you used something custom.

really thanks for the notebook!

[muellerzr (Zachary Mueller)]

@leaf

it looked like you used something custom.

It’ll be explained in next weeks WWF2

[leaf (jakub langr)]

awesome, btw I was looking at lesson5 and it still says Waiting for Zachary Mueller February 12, 2020 at 11:00 PM

Is that delayed then? No worries

[muellerzr (Zachary Mueller)]

Yes, if you check the most recent post life got busy, I posted when I’ll go live again for lesson 5 today

[NavneetSajwan]

Is the inference part ready yet??

[muellerzr (Zachary Mueller)]

@NavneetSajwan no, not fully. I have been busy with my college classes, the study group, and other responsibilities and could not get to it by last night. It’s partially working if you want to explore the code. It’s in inference.py. There will be bugs with it. And I’m unsure when exactly I will be able to work on it.

[NavneetSajwan]

Thanks, I will look into inference.py

[muellerzr (Zachary Mueller)]

@NavneetSajwan Fair warning: I won’t be able to debug anything for a few days to assist with it at all until I get a few other things done. It’s low on my totem pole.