Update readme

README.md

@@ -14,11 +14,11 @@ You can try a Streamlit demo app that uses this model on [🤗 Spaces](https://h
 ![Streamlit app](./assets/streamlit_app.png)
 
 🤗 Hub Model card: https://huggingface.co/flax-community/medclip-roco
-## Dataset 
+## Dataset 🧩
 
-Each image is accompanied by a text caption. The caption length varies from a few characters (a single word) to 2,000 characters. During preprocessing we remove all images that has a caption shorter than 10 characters.
+Each image is accompanied by a textual caption. The caption length varies from a few characters (a single word) to 2,000 characters (multiple sentences). During preprocessing we remove all images that has a caption shorter than 10 characters.
 Training set: 57,780 images with their caption.
-Validation set: 7.200
+Validation set: 7,200
 Test set: 7,650
 
 [ ] Give an example
@@ -26,7 +26,7 @@ Test set: 7,650
 ## Installation 💽
 This repo depends on the master branch of [Hugging Face - Transformers library](https://github.com/huggingface/transformers). First you need to clone the transformers repository and then install it locally (preferably inside a virtual environment) with `pip install -e ".[flax]"`.
 
-## Model
+## The Model ⚙️
 You can load the pretrained model from the Hugging Face Hub with
 ```
 from medclip.modeling_hybrid_clip import FlaxHybridCLIP
@@ -39,6 +39,13 @@ You can fine-tune a CLIP model implemented in Flax by simply running `sh run_med
 This is the validation loss curve we observed when we trained the model using the `run_medclip.sh` script.
 ![Validation loss](./assets/val_loss.png)
 
+## Limitations 🚨
+The current model is capable of identifying if a given radiology image is a PET scan or an ultrasound scan. However it fails at identifying a brain scan from a lung scan. ❗️This model **should not** be used in a medical setting without further evaluations❗️.
+
+## Acknowledgements
+Huge thanks to the Hugging Face 🤗 team and Google JAX/Flax team for organizing the community week and letting us use cloud compute for 2 weeks. We specially thank [@patil-suraj](https://github.com/patil-suraj) & [@patrickvonplaten](https://github.com/patrickvonplaten) for the continued support on Slack and the detailed feedback.
+
+
 ## TODO
 [ ] Evaluation on down-stream tasks