AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Long-Term Recurrent Convolutional Networks for Visual Recognition and Description

Jeff Donahue, Lisa Anne Hendricks, Marcus Rohrbach, Subhashini Venugopalan, Sergio Guadarrama, Kate Saenko, Trevor Darrell

TPAMI 2016 · Architectures and efficiency › RNNs and normalization · approach: LSTM design and memory · 5,239 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add LRCN LSTM on top of per-frame CNN (flow input)single-frame CNN · UCF101accuracy: 74.37 -> 77.28Improved
combine RGB and flow with weights 1/3, 2/3LRCN-fc6 · UCF101accuracy: 68.20 -> 82.34Improved
use LSTM instead of vanilla RNN (ablation)LRCN_2f CaffeNet · COCO val image-to-caption retrievalR@1: 30.2 -> 33.8Improved
use LSTM instead of vanilla RNN (ablation)LRCN_2f CaffeNet · COCO val captioningCIDEr-D: 0.703 -> 0.720Improved
factor the 2-layer LSTM (feed image only to layer 2) (ablation)LRCN 2-layer CaffeNet · COCO val caption-to-image retrievalR@1: 23.4 -> 25.0Improved
stack a second LSTM layer (unfactored vs 1 layer) (ablation)LRCN CaffeNet · COCO val caption-to-image retrievalR@1: 25.2 -> 23.4Hurt
fine-tune the CNN (ablation)LRCN_2f VGGNet · COCO val image-to-caption retrievalR@1: 46.0 -> 53.3Improved
fine-tune the CNN (ablation)LRCN_2f VGGNet · COCO val captioningCIDEr-D: 0.828 -> 0.896Improved
switch CaffeNet to VGGNet (ablation)LRCN_2f fine-tuned · COCO val captioningCIDEr-D: 0.764 -> 0.896Improved
sample best of N=100 (T=1.5) instead of greedy (ablation)LRCN_2f VGGNet FT · COCO val captioningCIDEr-D: 0.839 -> 0.896Improved
use beam width 3 instead of 1 (ablation)LRCN_2f VGGNet FT · COCO val captioningCIDEr-D: 0.839 -> 0.888Improved
increase beam width from 3 to 10 (ablation)LRCN_2f VGGNet FT · COCO val captioningCIDEr-D: 0.888 -> 0.886No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy