Share your work here ✅

I trained a model that takes a picture of a road accident, and tells us if its a car crash or a bike crash. The accuracy is about 93% with a resnet-34 model, before any data cleaning.

Most of the misclassified images had both, a car and a bike, involved in a crash.

After some data-cleaning (removing duplicate images, deleting unrelated images), I was able to improve the model performance to about 95%., where I used various learning rates to see what works best.

Here’s a reference notebook, if someone is interested:

Thanks to fastai team and the community :slight_smile:


Hi guys!
I am on lesson 2 and was wondering if I could use what I learnt on some ongoing problems. That is when I stumbled upon the idea of using a neural network to check if social distancing is being followed in a place or not…
There was no publicly available dataset so I scrapped Google for images on both the categories. Although I found a lot of images it was all mixed up. Apparently Google’s algorithm was not able to classify them correctly. :wink:
So I manually separated the images and trained a model. To my astonishment it was able to classify with 86% accuracy!

I think it can be used in CCTV or drones for surveillance of public places. It will certainly help law enforcement in these harsh pandemic.
You can check it out here:

If anyone is interested to collaborate please do, I have just started the project.

Special thanks to @joedockrill & @sachin93 for helping me setup the webapp(believe me or not it takes a lot more time and patience than training the network, if you are a beginner like me😅)


@mmd Amazing work! Glad to hear that I could be of help. I am a beginner myself and I understand your pain since I come from a Civil Engineering background with little programming experience.

I have released a small library to perform:

  • Knowledge Distillation
  • Sparsifying/Pruning of your models
  • Batch Normalization Folding
  • FC Layers factorization

All with fastai of course :slight_smile:

With these, I was able to halve the number of parameters and the inference time of a VGG16 model.

Blog post available here

And code available here


Hi vlad,

Could you add any any instructions how to do it, I am using the some configuration as you but I can’t install the whole thing.


Just built my version of the face-mask detector after completing lesson 2, with streamlit.
and here’s the repo.
Found all the help needed here in this forum. :smiling_face_with_three_hearts:


Another perspective of learning fastai course

I have been trying to do binary segmentation, for hair from peoples faces, and think its been doing quite well at the moment. I even combined it with a color LUT to do some nice recoloring of hair. !


@muellerzr this thread should be moved to 2020 i assume?

2020 has it’s own :slight_smile:

1 Like

Survival Analysis with pytorch using nbdev. At the end of the day it’s just another loss function. Made the docs so that (hopefully) you can understand the theory as well as code.

Still some work to be done in the Accelerated Failure Time models, but here’s the alpha:


I used UNET architecture for generators in Cycle GAN. I observed a significant improvement.

Here is my work

1 Like

I wrote a quick article on how to use Streamlit to quickly test your models. This can be used for fastai models as well. Link below

1 Like

Hi dipam7 hope all is well!
Great short post"
not registered with medium_com :clap:t2:
Cheers mrfabulous1 :smiley: :smiley:

1 Like


I have written a blog and developed a Covid19 chest X-ray image classifier based on the initial 3 lessons. Here are the links:


1 Like

I’ve figured out that we don’t have to do any fine-tuning but just correct the weights in the bigger kernel with respect to how much bigger it got.
5x5->9x9; have to multiply the weights by (5x5)/(9x9).

Nice mind map ! I like it

Hi everyone! Want to introduce the Weights & Biases report about transformers, which dives into the breakthroughs, scientific basis, formulas, and code for the transformer architecture, which you can read in the Chinese. :cn:
:woman_technologist:t2: Report:

Our paper using ULMFit to generate potential inhibitors of SARS-CoV-2 Mpro is finally here!


Recognising MNIST digits using mean of pixel values, without any machine learning algorithm

I have created a well documented python code for classifying digits without machine learning. Will be happy to hear the suggestions for improving the code.

I can reached at: