amartomar683 commited on
Commit
a4ff752
·
verified ·
1 Parent(s): 1c3b951

Clarify DINOv2-small is a frozen encoder, not a fine-tuned base model

Browse files
Files changed (1) hide show
  1. README.md +4 -2
README.md CHANGED
@@ -4,7 +4,6 @@ library_name: pytorch
4
  pipeline_tag: image-to-text
5
  datasets:
6
  - jxie/coco_captions
7
- base_model: facebook/dinov2-small
8
  tags:
9
  - image-captioning
10
  - video
@@ -22,7 +21,10 @@ description the assistant can read, through MCP (Claude Desktop, Claude Code, Cu
22
  VS Code, LM Studio and other MCP apps) or as plain text for any LLM. It runs locally, so
23
  nothing is uploaded.
24
 
25
- It's a **36M-parameter** image and video captioner, fast on a laptop CPU.
 
 
 
26
 
27
  - **89% of Florence-2-base's COCO caption score at 1/6 the size**
28
  - **8× faster on GPU, 75× faster on CPU** (54 ms per image on an Apple M5 CPU)
 
4
  pipeline_tag: image-to-text
5
  datasets:
6
  - jxie/coco_captions
 
7
  tags:
8
  - image-captioning
9
  - video
 
21
  VS Code, LM Studio and other MCP apps) or as plain text for any LLM. It runs locally, so
22
  nothing is uploaded.
23
 
24
+ It's a **36M-parameter** image and video captioner, fast on a laptop CPU. It uses Meta's
25
+ DINOv2-small as a frozen image encoder (its weights are unchanged, so this is not a
26
+ fine-tune of DINOv2); the adapter and caption decoder on top (14M parameters) were
27
+ trained from scratch on COCO.
28
 
29
  - **89% of Florence-2-base's COCO caption score at 1/6 the size**
30
  - **8× faster on GPU, 75× faster on CPU** (54 ms per image on an Apple M5 CPU)