Share your work here ✅

Hello everyone,

First of all, I want to thank fast.ai and the users for the amazing course and forum posts! Second, I would like to share a web application I made based on the second lecture. This is actually my first self-made project, and this course is the first in which I learn about machine learning. So I would like to receive some feedback.

The application I made is a melanoma classifier, which you can visit here: https://melanoma-classifier.herokuapp.com/. As you see from the url, I used Heroku to deploy the app. And you can find the source code here (including the Jupiter notebook with which I trained the model): https://github.com/Mees-Molenaar/melanoma-classifier.

In the notebook I have made lists of problems that occurred and possible improvements that can be made, which I will copy here:

Improvements

This is my second project (I am still finishing the first project). I already felt more comfortable using the fast.ai library. However, there is still room for improvement.

  • Making sure the training data is trustworthy since I am no dermatologist
  • Maybe using resnet50 instead of resnet34
  • Using other transforms (maybe just like the satellite in lesson 3, for this problem I can use vertical flips?)
  • And in lesson 3 there was a part about picture sizes, maybe that also can help this model. However, I didn’t fully understand it yet.
  • Improving the learning rate
  • Change the number of epochs.

Difficulties

During this project, I had several difficulties challenges that I would like to share with you.

  • Finding a trustworthy dataset (I was not able to find the source of these pictures)

  • I am no dermatologist so it is hard for me to verify the data.

  • Making the web-application. My only experience is using Flask in the cs50 course I followed online via Edx. However, the most problematic part was deploying the model on a hosting site. After a lot of tries, I was finally able to launch the app on Heroku.

  • I am not able to run the code locally yet, so I had to upload it to Heroku to check if it worked. This takes quite some time, so honestly, I should try to install the necessary programs locally. Since I don’t have NVIDIA graphics card, so don’t have CUDA, I for some reason was not able to install the fastai library (even using the pytorch CPU only library)

  • First time using Heroku: needed to change the Procfile, but I didn’t know how. I found out that for the web application you need this line of code:

web: gunicorn application:app

  • And you should rename application to your filename that creates the Flask instance, and app to the name you gives this Flask instance in your application file.
  • Lastly, I got a problem: Unexpected key(s) in state_dict. This seems to be a problem with different fastai library versions you use for learning the model and making predictions in your web app. So you have to make sure you use the same fastai library versions.

Thanks a lot for your time reading about this project, and I would be glad to answer any questions.

Cheers,

Mees

3 Likes

Congrats on your project! Just had a play with the app and it correctly predicted two examples (benign and not benign images).

I’ve been experimenting with creating image classifiers without requiring any code or data for that matter… I know it might be somewhat controversial, but I wanted to see how far I’d get…

I wrote down some context: https://www.zerotosingularity.com/blog/seeme-intro/ which also includes a video demo…

3 Likes

Hi, I create and put into Github two simple examples of monitors for fast.ai, one for Neptune.ml, and one for TensorBoard when you can use both in the Google Colab notebook. I hope that it’ll help use these two great tools.

The example of the Colab Notebook: https://colab.research.google.com/drive/1HEVbnOyKSLSIkMhpYLWO83Wut3hJrP42

List of saved parameters that are saved in the TensorBoard:

  • Scalars with all metrics and training loss for each batch
  • Images for top losses images
  • Graph for the model
  • Distribution and Histograms for each parameters in layers
  • Hyperparameter Text for learning rate (lr) , momentum (mom) , weight decay (wd) , and beta (beta)
  • Text for Model , Summary , Model save path , and Total Epochs

List of parameters that are saved in the Neptune.ml:

  • all metrics
  • last_loss batch
  • model.txt - info about the model
  • opt.txt - OptiWrapper
  • Properties for learning rate (lr) , momentum (mom) , weight decay (wd) , and beta (beta)
  • Use of CPU/GPU , RAM

Couple screenshots:

These files are straightforward, and you can modify for your purpose as you wish.

7 Likes

Hi all, I just completed the first week. I had a go at classifying painting styles using the dataset from the Painter by Numbers Kaggle competition (https://www.kaggle.com/c/painter-by-numbers/data).

I took the 20 most common styles from the training set (62k images) and ran them through the lesson notebook.

image

The loss seemed to stop improving after about 10 epochs (unfrozen), with accuracy just over the 60% mark, although it varied quite a lot by the style. The below is the confusion matrix, I used normalize = True since there were unequal numbers for each style.

Worst performing was ‘Art Informel’, which got very confused with ‘Abstract Expressionism’ - this made sense since Wikipedia says that they were developed in parallel, the main difference being one was from Europe and the other from America. The best performing was ‘Ukiyo-e’ (92% accuracy) which is paintings from the Japanese Edo period (1603 and 1868) - the isolationism of the age reflected by a particularly distinctive style perhaps.

Really enjoying the course, onwards to week 2…

4 Likes

Hi Guys

I just finished Lesson 1. Using the techniques taught in the lesson 1, I wrote 4 medium articles:

  1. Devanagari Handwritten Digits Classifier - In this challenge, I got an accuracy of 98.99%

  2. Malaria Blood Cell Images - In this challenge, I got an accuracy of 97.3%

  3. Pneumonia Chest X-ray images - In this challenge, I got an accuracy of 95.8%

  4. Histopathological Cancer Images - In this challenge, I got an accuracy of 98.12%

Comments and feedback shall be most welcome and appreciated.

I am simply amazed by mainly two things:

a. In few lines of codes and short span of training time, the level of accuracy is mind-boggling. Never knew that Deep Learning could be so much easier to implement on real-world challenges. A big thanks to Jeremy and Rachel for this wonderful course.

b. I used in all above different challenges same CNN architecture ResNet50 and just blown away by the fact that a model trained on a completely different dataset (ImageNet) is so accurate in classifying different types of unseen images.

Really looking forward to Lesson 2 now.

5 Likes

In an attempt to show my appreciation for Jeremy and also be able to hear him talk about whatever I wanted whenever I wanted, I made a Jeremy Howard language model I’ve been calling MiniJereme. I trained a couple different models because it kind of became too good so I tried making it worse to get some funnier responses. Here are some of its predictions:

I also thought this would be a good time to share a little about me since I don’t have to worry about Jeremy kicking me out of the course/ SF study group for lying about having one year of “coding” experience :sweat_smile:

I grew up playing baseball, played in college, majored in business management, and minored in philosophy and psychology. After I graduated I knew I didn’t want to play baseball anymore, ended up studying to go to law school for a while, then last April decided to look up how to program. Somehow I stumbled on fastai and found myself taking part 1 v3 in person and going to the SF study group everyday (when I got there I didn’t know how to do comprehensions).

Since then I’ve been able to apply the lessons to things like coloring a photo of my grandmas husband (who passed away before I was born) and giving it to her for Christmas, an app for my cousin (who plays professional baseball) that analyzes his swings for him, and speech recognition for my sister who has autism. It makes me so happy that I’ve been able to give my grandma memories she might not have thought of, or maybe make my sister’s or cousin’s lives better by designing things specifically for them!

I will always be thankful to Jeremy, Rachel, and Sylvain for everything I’ve learned and how you all have changed my life. Learning programming and fastai has let me get ideas out of my head that would otherwise be stuck and it’s an amazing feeling. I hope one day I can give back to you three what you have given to me :slightly_smiling_face: and a :wink: for Sylvain

18 Likes

firstly, i would like to say a very big thank you to Jeremy Howard and all the members of fastai for giving me the opportunity to pursue my dream. thank you.

secondly, i am currently offering the DL course by Jeremy lesson 2 to be precise, i can confidently build a state of the art vision model now but am having issues creating a model to solve the kaggle digit recognizer challenge. i believe the reason is because kaggle dataset is a tabular dataset not image. how can i go about it.

this is a great idea

2 Likes

All -

I wanted to share with you all an Art Project I have been working on: https://www.9gans.com/

9 GANS is An AI Generated Art Gallery that is refreshed every hour to create a completely new and unique collection of 9 images. Hence the (bad) name. Let me know what you think. :slight_smile:

17 Likes

After watching Jeremy’s examples of model interpretation in machine learning course (lesson 3 and lesson 4) I’ve wondered if something similar exists in tabular models. And looks like it didn’t, even though it seems to be very tempting to understand data through it’s model interpretation.
So I’ve implemented these functions and made an example for the Rossmann data.
There you can find functions for:

  • feature importance calculations
  • partial dependence calculation
  • plotting dendrograms (for data inself)
  • plotting embeddings
  • also there is an example of using trained embeddings for another models (using only, not retraining), like to compare NN with Random Forest+embeddings

The notebook is here

4 Likes

Hi @alvisanovari

I am also doing the similar project on Age Prediction but used instead 7GB IMDB-Wiki dataset.

By just looking at the data, it looks like a noisy dataset. Few of the celebs I know of have their ages completely way-off.

Not surprisingly, when I am using metric as “accuracy”, its not very great.

Do you have any idea if some other better alternative dataset available in public domain?

Best Regards
Abhik

Nice work! Have you seen the ELI5 library? It is somewhat similar to what you are doing, except it seems that it’s more focused on certain ML algorithms like XGBoost and LightGBM.

I didn’t go too deeply into this. You can try predicting buckets (5 year age bands) instead of exact ages to reduce some of the noise and get more ‘accurate labels’.

1 Like

Made a pokemon dataset and got 83% accuracy, guessing which pokemon there is on a picture. If you want to do something with pokemon images too please consider using my dataset: kaggle.com/mrgravelord/gen7-pokmon-image-dataset
It has more than 24k images and features almost every pokemon(except some forms like “Spiky-eared Pichu”).

I just launched Ask Roboflow, an AI trained to answer programming questions.

It’s the next step from my project last month, Stack Roboflow, which generated programming questions.

Give it a try, I’d love to hear your thoughts and feedback!

It’s learned lots of interesting things including how to insert HTML links and images into its replies, the syntax of several programming languages, and how to link to “relevant” documentation. And it even seems to be picking up a bit of a sense of humor; it replied “42” to one of my questions yesterday :laughing:

Unfortunately, one thing it hasn’t yet learned is the concept of “correctness” so most of the answers you’ll see won’t actually be helpful yet… I plan to continue to improve the model as time goes on. Hopefully one day it will actually be able to help new programmers get instant answers to their programming questions.


I’ve also published a Medium post about it. I’m thinking about following up with a post about the serverless architecture (Google Cloud Functions + Firebase) that I’m using if anyone is interested.

10 Likes

It will be helpful , if you could start a new thread for AI enabled question answering and post these there. Me too working on a similar task, will share details once i get respectable results.

1 Like

Great work Jason. Loved the way you used the learnings from fastai course to build all those apps. Keep going :clap::clap:

1 Like

Update on Crystal_Clear(My personnal project to upscale audio using image superresolution techniques). Since my last post 22 days ago, I had some time to make some progress on my project : https://github.com/StatisticDean/Crystal_Clear

  • Prediction is now parallelized and is 2-3 times faster. It takes around 0.5 sec to process a 30 sec audio on my GPU and around 60 sec if I use the slower processing method(Taking all overlapping windows instead of no overlapping windows).

  • I’ve created a Starter kit to process the data so you can do training yourself as well. At the moment processing the data is kinda slow(around 2-3 hours).

  • I’ve also worked with my brother to make a webapp.
    The Back Office repo can be found here : https://github.com/StarAurryon/Crystal_Clear-BO
    The webapp Link is the following : http://crystalclear.staraurryon.com
    You can upload an audio file (mp3, ogg or wav) which will be taken through the upscaling algorithm and can then be downloaded.

Please keep in mind that this is a work in progress, and that the algorithm might do nothing significant or even deteriorate the audio. The algorithm has been trained with audio encoded in mp3 with a very low bitrate, so I would suggest feeding it low quality mp3 to see interesting results. If you perform some tests with the webapp and get outstanding results, please contact me :slight_smile:

7 Likes

It is great to see people make so many projects here. I’ve been working on them too. Two things I’ve learnt during this period are:

  1. Execute code as much as possible

  2. Write about it.

The second one has helped me learn things better. It has forced me to read more about topics to get a clear understanding of the concepts and put it in a cogent way. It has also given me visibility. I’d encourage everyone to write about their work. Here are some of my recent posts:

Transfer learning in NLP

Lessons from my first deep learning hackathon

Mixed precision training

Collaborative filtering on song dataset

Data augmentation

Weight decay

Image regression

All kinds of feedback will be appreciated

2 Likes