FiLM: Visual Reasoning with a General Conditioning Layer

22 Sep 2017  ·  Ethan Perez, Florian Strub, Harm de Vries, Vincent Dumoulin, Aaron Courville ·

We introduce a general-purpose conditioning method for neural networks called FiLM: Feature-wise Linear Modulation. FiLM layers influence neural network computation via a simple, feature-wise affine transformation based on conditioning information. We show that FiLM layers are highly effective for visual reasoning - answering image-related questions which require a multi-step, high-level process - a task which has proven difficult for standard deep learning methods that do not explicitly model reasoning. Specifically, we show on visual reasoning tasks that FiLM layers 1) halve state-of-the-art error for the CLEVR benchmark, 2) modulate features in a coherent manner, 3) are robust to ablations and architectural modifications, and 4) generalize well to challenging, new data from few examples or even zero-shot.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Visual Question Answering (VQA) CLEVR CNN+GRU+FiLM Accuracy 97.7 # 10
Visual Question Answering (VQA) CLEVR-Humans CNN+GRU+FiLM Accuracy 75.9 # 3
Image Retrieval with Multi-Modal Query MIT-States FiLM Recall@1 10.1 # 4
Recall@5 27.7 # 4
Recall@10 38.3 # 5

Methods


No methods listed for this paper. Add relevant methods here