MearcAll papers

arXiv 1608.06993 · cs.CV cs.LG

Densely Connected Convolutional Networks

Gao Huang, Zhuang Liu, Laurens van der Maaten, Kilian Q. Weinberger

Claims were extracted automatically by reducer-1 on 2026-11-12. The paper's license does not allow republishing it, so long claims are shortened to a quotation. Read the full text on arXiv.

No reproduction recordedclm_3594e992
DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less computation to achieve high performance.

Abstract · page 1 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_659164ad
1 arXiv:1608.06993v5 [cs.CV] 28 Jan 2018 In this paper, we propose an architecture that distills this insight into a simple connectivity pattern: to ensure maximum information flow between layers in … (excerpt)

Introduction · page 1 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedclm_8605cb0b
Further, we significantly outperform the current state-ofthe-art results on most of the benchmark tasks.

Introduction · page 2 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedclm_2ff1872d
To further improve the information flow between layers we propose a different connectivity pattern: we introduce direct connections from any layer to all subsequent layers.

DenseNets · page 3 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedclm_9c106c9e
Note that each “conv” layer shown in the table corresponds the sequence BN-ReLU-Conv. obtain state-of-the-art results on the datasets that we tested on.

DenseNets · page 4 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedclm_1bf37010
To highlight general trends, we mark all results that outperform the existing state-of-the-art inboldface and the overall best result in blue.

The test errors were only evaluated once for each task · page 6 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_e31de8f3
Possibly the most noticeable trend may originate from the bottom row of Table 2, which shows that DenseNet-BC with L = 190 and k = 40 outperforms the existing state-of-the-art … (excerpt)

The test errors were only evaluated once for each task · page 6 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_255ef9a7
Our best results on C10 and C100 (without data augmentation) are even more encouraging: both are close to 30% lower than Fractal- Net with drop-path regularization [17].

The test errors were only evaluated once for each task · page 6 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_ab079de7
On SVHN, with dropout, the DenseNet with L = 100 and k = 24 also surpasses the current best result achieved by wide ResNet.

The test errors were only evaluated once for each task · page 6 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_3efacceb
However, the 250-layer DenseNet-BC doesn’t further improve the performance over its shorter counterpart.

The test errors were only evaluated once for each task · page 6 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_1beb2c91
For example, our 250-layer model only has 15.3M parameters, but it consistently outperforms other models such as FractalNet and Wide ResNets that have more than 30M parameters.

The test errors were only evaluated once for each task · page 6 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_40baf50b
We also highlight that DenseNet-BC with L = 100 andk = 12 achieves comparable performance (e.g., 4.51% vs 4.62% error on C10+, 22.27% vs 22.71% error on C100+) as the … (excerpt)

The test errors were only evaluated once for each task · page 6 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_ea42285e
The 1001-layer deep ResNet converges to a lower training loss value but a similar test error.

The test errors were only evaluated once for each task · page 6 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_2f2a2ee2
On C10, the improvement denotes a 29% relative reduction in error from 7.33% to 5.19%.

The test errors were only evaluated once for each task · page 6 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_ec08ce07
DenseNet-BC requires about 1/3 of the parameters as ResNet to achieve comparable accuracy.

The test errors were only evaluated once for each task · page 7 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedres_bc777cee
Further, to achieve the same level of accuracy, DenseNet-BC only requires around 1/3 of the parameters of ResNets (middle plot).

Discussion · page 7 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedclm_cb5fed73
Under multiple settings, it achieved state-of-the-art results across several highly competitive datasets.

Conclusion · page 8 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedclm_eefc2c3a
Moreover, DenseNets require substantially fewer parameters and less computation to achieve state-of-the-art performances.

Conclusion · page 8 · Automated extraction, not reviewed · 0 reviewed reproductions

No reproduction recordedclm_dc23903f
Because we adopted hyperparameter settings optimized for residual networks in our study, we believe that further gains in accuracy of DenseNets may be obtained by more detailed tuning of hyperparameters … (excerpt)

Conclusion · page 8 · Automated extraction, not reviewed · 0 reviewed reproductions

Reproduced one of these claims?

Record your result in the Mearc app on your own machine, then use Publish this result with this paper's id (1608.06993) and the claim id shown next to the claim. Mearc creates a file that holds only what you choose to publish. Add it to the public claim data by pull request; a maintainer other than you reviews it before it counts.