メニュー
Render stroke sequences to grayscale images with temporal intensity encoding where earlier strokes are brighter and later strokes fade to encode drawing order
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | cv-stroke-temporal-color-rendering |
| description | Render stroke sequences to grayscale images with temporal intensity encoding where earlier strokes are brighter and later strokes fade to encode drawing order |
Hand-drawn sketches are stored as ordered stroke sequences. When rendering to an image, encoding the drawing order as pixel intensity — earlier strokes brighter, later strokes darker — gives CNNs an additional temporal signal beyond pure geometry. This simple trick improves doodle classification accuracy by 1-3% over flat-color rendering.
import cv2
import numpy as np
import json
def draw_strokes(raw_strokes, size=64, line_width=6):
img = np.zeros((256, 256), np.uint8)
for t, stroke in enumerate(raw_strokes):
color = max(255 - t * 13, 30) # fade from 255 to ~30
for i in range(len(stroke[0]) - 1):
cv2.line(img,
(stroke[0][i], stroke[1][i]),
(stroke[0][i + 1], stroke[1][i + 1]),
color, line_width)
if size != 256:
img = cv2.resize(img, (size, size))
return img
strokes = json.loads(drawing_string)
img = draw_strokes(strokes, size=64)
[[x0, x1, ...], [y0, y1, ...]] per stroke255 - t * step, clamped to a minimum13 per stroke works for ~20-stroke drawings; adjust for longer sequencesMemory-efficient Keras generator that streams sharded CSV files in chunks, renders strokes to images on-the-fly, and yields batches for training on datasets too large for memory
Stack 1D convolutions for local feature extraction before bidirectional LSTMs to classify variable-length stroke sequences into hundreds of doodle categories
Partition a large dataset into N balanced shards using integer key modulo arithmetic for reproducible, class-interleaved splits across CSV files
Normalize raw stroke coordinates to 0-255 range, resample at uniform arc-length spacing, then apply Ramer-Douglas-Peucker simplification
Score each face in a video as anomalous by computing L2 distance from the embedding centroid of all faces, then convert to probability via logistic function
Extract per-frame CNN features then classify the temporal sequence with stacked GRU layers and a boolean mask for variable-length video inputs