No description, website, or topics provided.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

Cascaded Mutual Modulation for Visual Reasoning

Source code for Yiqun Yao; Jiaming Xu and Bo Xu. Cascaded Mutual Modulation for Visual Reasoning. 2018. EMNLP. arxiv

Code Outline

This code is a fork from the code for "FiLM: Visual Reasoning with a General Conditioning Layer" available here.

We implement a new model: CMM on the basis of PG+EE and FiLM. Different from the original code, our model runs on multiple gpus.

Setup and Training

Setup instructions for the CMM model are nearly the same as for PG+EE and FiLM.

First, follow the virtual environment setup instructions.

Second, follow the CLEVR data preprocessing instructions.

The below script has the hyper-parameters and settings to reproduce CMM CLEVR results:

sh scripts/train/

Running models

The following scripts run trained models on CLEVR:

python scripts/