Unified Model of Hippocampal Spatial and Object Cells

Overview

A unified computational model that simulates the emergence of spatial and object-sensitive neurons in the hippocampus. By integrating visual inputs (LEC) and path integration (MEC) through a Graph Neural Network, the model reproduces seven cell types observed experimentally: place cells, grid cells, border cells, object cells, object-sensitive cells, object-vector cells, and object-trace cells.

Architecture

• Visual branch: CNN encoder processing 32×32 agent-POV frames
• Path integration branch: Dense layer on 100-dim trajectory vectors
• Integration: Graph Convolutional Network (GNN) connecting LEC and MEC streams
• Output heads: Multi-task prediction of position (22), heading (11), and reward (1)

Deployment

The model is served via a FastAPI REST endpoint and interactive Streamlit demo, both deployed on HuggingFace Spaces.

• API: POST /inference — upload trajectory + image files, returns firing rate maps as base64 images
• Demo: Upload .pk1 trajectory files and visualise neuron firing maps by layer

Repository Structure

├── notebooks/
│   └── simulations/
│       ├── common_model_ver_no.ipynb  # Model architecture and support functions
│       └── main.ipynb                 # Main simulation notebook
├── matlab_support_files/
│   └── get_firing_rates.m             # Shuffling tests for grid cells
├── download_data.py                   # Downloads weights and data from HuggingFace
├── requirements.txt
└── README.md

Getting Started

1. Clone the repository

git clone https://github.com/jarvez31/Object_representation_model.git
cd Object_representation_model

2. Install dependencies

pip install -r requirements.txt

3. Download model weights and data

python download_data.py

4. Run the notebooks

Open notebooks/simulations/main.ipynb and follow the instructions.

Results

Cell Type	Emerged
Place cells	✅
Grid cells	✅
Border cells	✅
Object cells	✅
Object-sensitive cells	✅
Object-vector cells	✅
Object-trace cells	✅

Modelling Perspective

Experimental Results

Multisensory Integration Model

Model Results

Citation

@article{patil2024unified,
  title={A unified model of hippocampal spatial and object cells},
  author={Patil, Bharat K.},
  journal={bioRxiv},
  year={2024},
  doi={10.1101/2024.09.09.612040}
}