Research map / Vision-language and multimodal / Human and embodied modeling
Regularizing RNNs for Caption Generation by Reconstructing the Past With the Present
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Add ARNet to the attentive encoder-decoder | Attentive Encoder-Decoder · MSCOCO Karpathy test split | CIDEr: 0.996 -> 1.034 | Helped |
| Add ARNet to the attentive encoder-decoder | Attentive Encoder-Decoder · HabeasCorpus testing split | BLEU-4: 0.088 -> 0.120 | Helped |
| Add ARNet to the LSTM | LSTM · Permuted sequential MNIST | Test Accuracy: 0.914 -> 0.933 | Helped |
| Add ARNet to the attentive encoder-decoder | Attentive Encoder-Decoder · MSCOCO image captioning | mean centroid distance: 0.773 -> 0.491 | Helped |
| Add ARNet to the attentive encoder-decoder | Attentive Encoder-Decoder · HabeasCorpus code captioning | mean centroid distance: 0.594 -> 0.322 | Helped |
| Increase the ARNet loss weight too much (ablation) | Attentive Encoder-Decoder + ARNet · MSCOCO image captioning | CIDEr | Hurt |
| Use a positive ARNet loss weight instead of lambda = 0 (ablation) | Attentive Encoder-Decoder + ARNet · MSCOCO image captioning | CIDEr | Helped |
| Add zoneout instead of leaving the encoder-decoder unregularized | Encoder-Decoder · MSCOCO Karpathy test split | CIDEr: 0.961 -> 0.941 | Hurt |
| Add zoneout instead of leaving the attentive encoder-decoder unregularized | Attentive Encoder-Decoder · MSCOCO Karpathy test split | CIDEr: 0.996 -> 0.975 | Hurt |
Closest papers by meaning
- Review Networks for Caption Generation (NeurIPS 2016 · similarity 0.88)
- Reconstruction Network for Video Captioning (CVPR 2018 · similarity 0.88)
- Improving Image Captioning with Conditional Generative Adversarial Nets (AAAI 2019 · similarity 0.87)
- Recurrent Nested Model for Sequence Generation (AAAI 2020 · similarity 0.87)
- Incorporating Copying Mechanism in Image Captioning for Learning Novel Objects (CVPR 2017 · similarity 0.86)
- Temporal Deformable Convolutional Encoder-Decoder Networks for Video Captioning (AAAI 2019 · similarity 0.86)
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.