Update readme

README.md

@@ -1,7 +1,7 @@
 # Tiger classification
 
-This repository contains scripts and notebooks to classify tigers (and other species) in camera trap images,
-using ML (e. g. [MegaDetector](https://github.com/agentmorris/MegaDetector)),
+This repository contains scripts and notebooks to build a model that can classify tigers (and other species) in camera trap images,
+using ML (e. g. [MegaDetector](https://github.com/agentmorris/MegaDetector) and [MEWC](https://github.com/zaandahl/mewc)),
 open source tools and data (e. g. [LILA BC](https://lila.science/))
 and free compute resources (i. e. Colab and Kaggle).
 
@@ -28,32 +28,52 @@ and make it available through [EcoAssist](https://addaxdatascience.com/ecoassist
 - [LILA BC](https://lila.science/)
 - amur tiger re-identification [challenge](https://cvwc2019.github.io/challenge.html) at CVWC 2019
 
-**Sample and download images (Colab)**
+**Sample and download images**
+
+[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/alexvmt/tiger_classification/blob/main/notebooks/sample_and_download_images.ipynb)
 
 1. Download image URLs and labels from LILA BC
 2. For each selected species: sample and download images, create train test split if applicable
 3. Copy images to Drive
 
 *Note: Since Colab and Drive have limited capacities, one might have to further split up the process.*
+*Note: I found the image downloading to be much faster in Colab and Drive compared to Kaggle.*
+
+**Preprocess images**
 
-**Preprocess images (Kaggle)**
+[Open in Kaggle](https://www.kaggle.com/code/alexvmt/preprocess-images/notebook)
 
-1. Run [MegaDetector](https://github.com/agentmorris/MegaDetector) on all images
-2. Snip images
+1. Run MegaDetector on all images
+2. Snip images [following mewc-snip](https://github.com/zaandahl/mewc-snip)
 3. Copy snipped images to Kaggle Output
 
 *Note: Images must have been previously downloaded to Drive via Colab and then uploaded to Kaggle (zipped folder).*
+*Note: I found access to free GPUs much better and transparent in Kaggle compared to Colab.*
 
 ## Training
 
-- [MEWC](https://github.com/zaandahl/mewc)
-- EfficientNetV2
-- Kaggle
+[Open in Kaggle](https://www.kaggle.com/code/alexvmt/training/notebook)
+
+1. Use [Keras Image Models](https://github.com/james77777778/keras-image-models)
+2. Follow [mewc-train](https://github.com/zaandahl/mewc-train)
+3. Log experiments using [Weights & Biases](https://wandb.ai/alexvmt/tiger_classification/overview)
+
+I selected a pre-trained EfficientNetV2S with 21 mio parameters because it constitutes a good compromise between predictive performance, training time and model size.
+The model has been trained for 30 epochs (early stopping after 24 epochs) with 4000 images per class.
+The model has been evaluated on 300 images per class. Below is the resulting confusion matrix.
+
+![confusion_matrix](media/confusion_matrix.jpg 'confusion_matrix')
+
+Other metrics can be found in the [respective experiment run](https://wandb.ai/alexvmt/tiger_classification/runs/0an7w90t/overview) on Weights & Biases.
+
+*Note: There are only ~300 tiger images on LILA BC. I didn't use them in training but instead put all of them in `test2`
+to examine how the model would potentially generalize to tiger camera trap images from another source than the tiger training images
+(like it would be the case with the Nepal Tiger Trust using the model on their own images through EcoAssist).*
 
 ## Deployment
 
-- HuggingFace
-- [EcoAssist](https://addaxdatascience.com/ecoassist/)
+1. Publish model on [HuggingFace](https://huggingface.co/alexvmt/tiger_classification/tree/main)
+2. Integrate and use model in [EcoAssist](https://addaxdatascience.com/ecoassist/)
 
 Join [AI for Conservation Slack](https://beerys.github.io/#slack) and [WILDLABS](https://wildlabs.net/) if you're interested in using technology for conservation.