SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1905.12681 · 2019

What Makes Training Multi-Modal Classification Networks Hard?

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 11 functions out of this paper's own repositories and ran 5 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
facebookresearch/R2Plus1D canonical 5 of 11
FunctionStatusWhere it lives
accuracy_metric Ran facebookresearch/R2Plus1D/c2/lib/utils/metric.py
code served (permissive licence) · get_code("eafa75d1a397bede")
blobs_from_pkl Ran facebookresearch/R2Plus1D/utilities/model_conversion/convert_models.py
code served (permissive licence) · get_code("be8f28f90ea3b6fd")
create_model Ran facebookresearch/R2Plus1D/c2/lib/models/c3d_model.py
code served (permissive licence) · get_code("33fb556271040441")
mean_ap_metric Ran facebookresearch/R2Plus1D/c2/lib/utils/metric.py
code served (permissive licence) · get_code("3f1d35c210ec7379")
model_validation Ran facebookresearch/R2Plus1D/c2/lib/models/model_builder.py
code served (permissive licence) · get_code("4a0058a818794147")
FlipBGR2RGB Not yet run facebookresearch/R2Plus1D/c2/lib/utils/model_loader.py
code served (permissive licence) · get_code("dfc355d0820eea25")
ir_csn_152 Not yet run facebookresearch/R2Plus1D/utilities/model_conversion/conversion_models.py
code served (permissive licence) · get_code("3c5043c8ffebe8fc")
r2plus1d_152 Not yet run facebookresearch/R2Plus1D/pt/vmz/models/r2plus1d.py
code served (permissive licence) · get_code("f5d71fa38ad9ad76")
r2plus1d_152 Not yet run facebookresearch/R2Plus1D/utilities/model_conversion/conversion_models.py
code served (permissive licence) · get_code("f6417808cd2990b8")
r2plus1d_34 Not yet run facebookresearch/R2Plus1D/pt/vmz/models/r2plus1d.py
code served (permissive licence) · get_code("1793f9b13efb9ae4")
r2plus1d_34 Not yet run facebookresearch/R2Plus1D/utilities/model_conversion/conversion_models.py
code served (permissive licence) · get_code("c2d08c4f7298be18")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Consider end-to-end training of a multi-modal vs. a single-modal network on a task with multiple input modalities: the multi-modal network receives more information, so it should match or outperform its single-modal counterpart. In our experiments, however, we observe the opposite: the best single-modal network always outperforms the multi-modal network. This observation is consistent across different combinations of modalities and on different tasks and benchmarks. This paper identifies two main causes for this performance drop: first, multi-modal networks are often prone to overfitting due to increased capacity. Second, different modalities overfit and generalize at different rates, so training them jointly with a single optimization strategy is sub-optimal. We address these two problems with a technique we call Gradient Blending, which computes an optimal blend of modalities based on their overfitting behavior. We demonstrate that Gradient Blending outperforms widely-used baselines for avoiding overfitting and achieves state-of-the-art accuracy on various tasks including human action recognition, ego-centric action recognition, and acoustic event detection.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("1905.12681")
get_code_for_paper("1905.12681")
have("1905.12681")

Connect an agent — have() is free.