Research map / Vision-language and multimodal / CLIP and zero-shot
Learning Deep Representations of Fine-Grained Visual Descriptions
Check 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 |
|---|---|---|---|
| Train both image-to-text and text-to-image objectives (DS-SJE rather than DA-SJE) (ablation) | Word-CNN-RNN · CUB | Top-1 Acc (%): 54.3 -> 56.8 | Helped |
| Train both image-to-text and text-to-image objectives (DS-SJE rather than DA-SJE) (ablation) | Word-CNN-RNN · CUB | AP@50 (%): 4.8 -> 48.7 | Helped |
| Train both image-to-text and text-to-image objectives (DS-SJE rather than DA-SJE) (ablation) | Word-CNN-RNN · Flowers | AP@50 (%): 7.6 -> 59.6 | Helped |
| Add a recurrent module above the character CNN (ablation) | Character-level text encoder · CUB | Top-1 Acc (%): 48.2 -> 54.0 | Helped |
| Use character CNN-RNN instead of character LSTM (ablation) | Character-level text encoder · CUB | Top-1 Acc (%): 21.6 -> 54.0 | Helped |
| Use word CNN-RNN instead of word LSTM (ablation) | Word-level text encoder · CUB | Top-1 Acc (%): 53.0 -> 56.8 | Helped |
| Use word CNN-RNN instead of word LSTM (ablation) | Word-level text encoder · Flowers | Top-1 Acc (%): 64.5 -> 65.6 | Helped |
| Use word CNN-RNN instead of word LSTM (ablation) | Word-level text encoder · Flowers | AP@50 (%): 52.3 -> 59.6 | Mixed |
| Train a high-capacity neural text encoder rather than using attributes | DS-SJE, Word-CNN-RNN · CUB | Top-1 Acc (%): 50.4 -> 56.8 | Helped |
| Use a word-level CNN-RNN rather than bag-of-words | DS-SJE · Flowers | Top-1 Acc (%): 57.7 -> 65.6 | Helped |
| Use Word-CNN-RNN rather than BoW for a single text query | Zero-shot retrieval · CUB, single query sentence | AP@50 (%): 14.6% -> 20.7% | Helped |
| Increase the number of training descriptions per image (ablation) | Neural text encoders · CUB | Top-1 classification accuracy | Helped |
Closest papers by meaning
- Improved Zero-Shot Classification by Adapting VLMs with Text Descriptions (CVPR 2024 · similarity 0.90)
- Follow-Up Differential Descriptions: Language Models Resolve Ambiguities for Image Classification (ICLR 2024 · similarity 0.87)
- Tree of Attributes Prompt Learning for Vision-Language Models (ICLR 2025 · similarity 0.87)
- Democratizing Fine-grained Visual Recognition with Large Language Models (ICLR 2024 · similarity 0.87)
- Class Semantic Attribute Perception Guided Zero-Shot Learning (AAAI 2025 · similarity 0.87)
- Analyzing and Boosting the Power of Fine-Grained Visual Recognition for Multi-modal Large Language Models (ICLR 2025 · similarity 0.87)
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.