AGENT SKILL

automatic-mask-generation

Generate masks for all objects in images or folders with Segment Anything's SamAutomaticMaskGenerator and the bundled AMG CLI. Use for batch automatic masks, PNG/CSV outputs, COCO RLE JSON, threshold tuning, crop settings, and memory-aware AMG runs.

License 许可
Apache-2.0
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
automatic-mask-generation
license
Apache 2.0
metadata
{ "disco-role": "operating" }
description
Generate masks for all objects in images or folders with Segment Anything's SamAutomaticMaskGenerator and the bundled AMG CLI. Use for batch automatic masks, PNG/CSV outputs, COCO RLE JSON, threshold tuning, crop settings, and memory-aware AMG runs.
disable-model-invocation
true

02 资源文件 5

03 SKILL.md 正文

Automatic Mask Generation

Use this sub-skill when the user wants SAM to segment all visible objects in an image or folder without point or box prompts. It covers SamAutomaticMaskGenerator, the bundled scripts/amg_cli.py, binary mask PNG folders, metadata.csv, COCO RLE JSON, threshold tuning, crop expansion, batching, and memory tradeoffs.

Do not use this sub-skill for prompted point/box/mask refinement; route those requests to ../prompted-segmentation/. Do not use it for ONNX export, browser inference, or the web demo; route those requests to ../onnx-and-browser/.

Quick Start

python sub-skills/automatic-mask-generation/scripts/amg_cli.py \
  --checkpoint sam_vit_b_01ec64.pth \
  --model-type vit_b \
  --input images/ \
  --output masks/ \
  --device cpu

For COCO-style RLE JSON instead of per-mask PNG folders:

python sub-skills/automatic-mask-generation/scripts/amg_cli.py \
  --checkpoint sam_vit_h_4b8939.pth \
  --model-type vit_h \
  --input image.jpg \
  --output masks-rle/ \
  --convert-to-rle \
  --device cuda

Routing

  • Use references/api-reference.md for direct Python use of SamAutomaticMaskGenerator and returned annotation records.
  • Use references/cli-reference.md for exact bundled CLI flags, output layout, and optional dependency checks.
  • Use references/workflows.md for folder runs, COCO RLE conversion, threshold tuning, and avoiding GPU out-of-memory failures.
  • Use references/troubleshooting.md for missing cv2, missing pycocotools, checkpoint/model mismatch, CPU fallback, unreadable images, empty output, and memory blowups.

Key Defaults

  • Registry keys are default, vit_h, vit_l, and vit_b; default is equivalent to the ViT-H builder.
  • SamAutomaticMaskGenerator(model) defaults to points_per_side=32, points_per_batch=64, pred_iou_thresh=0.88, stability_score_thresh=0.95, crop_n_layers=0, min_mask_region_area=0, and output_mode="binary_mask".
  • output_mode="coco_rle" requires pycocotools; min_mask_region_area > 0 requires OpenCV.
  • Large images, high points_per_side, large points_per_batch, crop layers, and binary mask output all increase memory use.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-triton-opt

Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-solve

END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-skill

PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-pattern-fusion

Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-orchestrate

Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-feedback

Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).

★ 351