grounding-dino

Name: grounding-dino
Author: graph-robots

26stars4forks

How to install grounding-dino?

graph-robots/open-robot-skills/grounding-dino

$npx -y skills add graph-robots/open-robot-skills --skill grounding-dino

Installs into the current project.

Summary

Grounding DINO zero-shot object detection — natural-language

Files · 1

View on GitHub

SKILL.md

1	# grounding-dino
2
3	The Grounding DINO servicer (`IDEA-Research/grounding-dino-base` via
4	transformers) as one in-process tool. Image in: RGB uint8 `[H, W, 3]` numpy
5	array; out: `{detections: [{box, label, score}, ...]}`.
6
7	## When to use
8
9	- Locating a named object in a camera frame: `grounding-dino.detect(rgb,
10	"cream cheese box.")`, pick the best box (highest score, or closest to a
11	pointing-model pixel), then `sam3.segment_box` for a pixel-accurate mask.
12	- Empty `detections` means nothing cleared the thresholds — treat as
13	not-found, don't retry blindly with the same prompt.
14
15	## Install
16
17	```bash
18	uv sync --extra grounding-dino # torch + transformers
19	# (pip: pip install -e ".[grounding-dino]")
20	```
21
22	Weights download from HuggingFace on first call. Env knobs:
23	`GAP_DINO_DEVICE` (default `cuda`; CPU works but is slow) and
24	`GAP_DINO_MODEL` (default `IDEA-Research/grounding-dino-base`).
25
26	## Gotchas (carried over from the servicer)
27
28	- Period-separated phrases: GDINO's text encoder expects each object
29	phrase terminated with `.` (`"red cube. green cube."`). A missing final
30	period is appended automatically, but separate multiple objects yourself.
31	- Default thresholds are deliberately low (0.20/0.20) for recall on
32	household objects; raise them when false positives leak through. Zero or
33	negative thresholds fall back to the defaults (proto-default semantics).
34	- `label` strings are the matched text spans, not your full query — when
35	querying multiple phrases, group detections by label.
36	- The model is a lazy module-level singleton; the first call pays the
37	weights-load latency, subsequent calls don't.

People who liked this also viewed

6 picks

Type

find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 20262.6M27k
grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026633k182k
vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026573k29k
improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026522k182k
setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026454k182k
triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026424k182k

grounding-dino

26stars4forks

View on GitHub

How to install grounding-dino?

graph-robots/open-robot-skills/grounding-dino

$npx -y skills add graph-robots/open-robot-skills --skill grounding-dino

Installs into the current project.

Summary

Grounding DINO zero-shot object detection — natural-language

Files · 1

View on GitHub

SKILL.md

1	# grounding-dino
2
3	The Grounding DINO servicer (`IDEA-Research/grounding-dino-base` via
4	transformers) as one in-process tool. Image in: RGB uint8 `[H, W, 3]` numpy
5	array; out: `{detections: [{box, label, score}, ...]}`.
6
7	## When to use
8
9	- Locating a named object in a camera frame: `grounding-dino.detect(rgb,
10	"cream cheese box.")`, pick the best box (highest score, or closest to a
11	pointing-model pixel), then `sam3.segment_box` for a pixel-accurate mask.
12	- Empty `detections` means nothing cleared the thresholds — treat as
13	not-found, don't retry blindly with the same prompt.
14
15	## Install
16
17	```bash
18	uv sync --extra grounding-dino # torch + transformers
19	# (pip: pip install -e ".[grounding-dino]")
20	```
21
22	Weights download from HuggingFace on first call. Env knobs:
23	`GAP_DINO_DEVICE` (default `cuda`; CPU works but is slow) and
24	`GAP_DINO_MODEL` (default `IDEA-Research/grounding-dino-base`).
25
26	## Gotchas (carried over from the servicer)
27
28	- Period-separated phrases: GDINO's text encoder expects each object
29	phrase terminated with `.` (`"red cube. green cube."`). A missing final
30	period is appended automatically, but separate multiple objects yourself.
31	- Default thresholds are deliberately low (0.20/0.20) for recall on
32	household objects; raise them when false positives leak through. Zero or
33	negative thresholds fall back to the defaults (proto-default semantics).
34	- `label` strings are the matched text spans, not your full query — when
35	querying multiple phrases, group detections by label.
36	- The model is a lazy module-level singleton; the first call pays the
37	weights-load latency, subsequent calls don't.

People who liked this also viewed

6 picks

Type