AGENT SKILL

sampling-and-indexing

Standardize video sampling and frame indexing so interval instructions and mask frames stay aligned with a valid key/index scheme.

License 许可
Apache-2.0
最近更新
2026-11-14
Stars
★ 0
Forks
1

01 前言 Front Matter

name
sampling-and-indexing
description
Standardize video sampling and frame indexing so interval instructions and mask frames stay aligned with a valid key/index scheme.

03 SKILL.md 正文

When to use

  • You need to decide a sampling stride/FPS and ensure all downstream outputs (interval instructions, per-frame artifacts, etc.) cover the same frame range with consistent indices.

Core steps

  • Read video metadata: frame count, fps, resolution.
  • When an exact target FPS is requested, sample on one regular time grid beginning at t=0. Endpoint inclusion is subordinate to that cadence: never append source frame n-1 merely to include the physical last frame. Include it only when the regular target-FPS grid selects it.
  • Keep the two index spaces separate: sample_ids[k] is a source-video frame ID, while k is the dense sampled-frame ordinal used by all output keys.
  • For N = len(sample_ids), per-frame artifacts must be exactly f_0 ... f_{N-1}. Half-open instruction intervals use 0 <= start < end <= N and must cover sampled ordinals 0 ... N-1 exactly once.

Pseudocode

import cv2, numpy as np
VIDEO_PATH = "<path/to/video>"
cap=cv2.VideoCapture(VIDEO_PATH)
n=int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
src_fps=cap.get(cv2.CAP_PROP_FPS)
target_fps=5.0
last_t=(n-1)/src_fps
k_max=int(np.floor(last_t*target_fps+1e-9))
sample_times=np.arange(k_max+1, dtype=float)/target_fps
sample_ids=np.rint(sample_times*src_fps).astype(int)
sample_ids=np.unique(sample_ids[(sample_ids>=0) & (sample_ids<n)]).tolist()
# Do not append n-1 unless it is already selected by this cadence.
N=len(sample_ids)
# Generate downstream keys with ordinals 0..N-1, not source IDs.

Self-check list

  • [ ] sample_ids strictly increasing, all < total frame count.
  • [ ] Consecutive sample times follow the requested cadence; no off-grid endpoint was appended.
  • [ ] JSON keys are plain start->end, use sampled ordinals, and their half-open intervals cover exactly 0..N-1 with no gap or extra ordinal.
  • [ ] Any per-frame artifact store contains exactly N consecutive sampled ordinals and no extras.

04 同仓库更多 Skills