Video Chaptering on VidChapters-7M Long (test)
41.3F1 ScoreGemini-1.5-Pro
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Gemini-1.5-ProBackbone=Gemini-1.5-Pro, Frame selection=Speech-based, Finetuned=false2025.03 | 41.3 | 70.6 | 10.1 | 55.3 | |
| Chapter-LlamaBackbone=Llama 3.1-8B, Frame selection=Proposed, Finetuned=true2025.03 | 41.3 | 69.2 | 15.8 | 91.2 | |
| GPT-4oBackbone=GPT-4o, Frame selection=Speech-based, Finetuned=false2025.03 | 36.5 | 66.2 | 6.6 | 34.8 | |
| Gemini-2.0-FlashBackbone=Gemini-2.0-Flash, Frame selection=Speech-based, Finetuned=false2025.03 | 34.9 | 66.2 | 9 | 51.6 | |
| GPT-4o-miniBackbone=GPT-4o-mini, Frame selection=Speech-based, Finetuned=false2025.03 | 28 | 61 | 6 | 27.3 | |
| Llama 3.1-8BBackbone=Llama 3.1-8B, Frame selection=Speech-based, Finetuned=false2025.03 | 26.6 | 59.3 | 3.6 | 18.9 | |
| Vid2SeqBackbone=Vid2Seq, Frame selection=Equidistant, Finetuned=true2025.03 | 16.7 | 50.8 | 5.9 | 28.4 | |
| Vid2SeqBackbone=Vid2Seq, Frame selection=Equidistant, Finetuned=false2025.03 | 4.6 | 32 | 0.3 | 0.5 |