3467dbcf03
Full-stack application for predicting where humans look in images using DeepGaze saliency models. Includes heatmap overlays, gaze sequence prediction, hotspot detection, AOI analysis, rule-based insights, optional Claude AI design analysis, and professional PDF report generation. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
47 lines
1.3 KiB
Python
47 lines
1.3 KiB
Python
import numpy as np
|
|
|
|
|
|
def compute_aoi_attention(
|
|
saliency: np.ndarray,
|
|
regions: list[dict],
|
|
) -> list[dict]:
|
|
"""
|
|
Compute attention metrics for user-defined regions.
|
|
|
|
Returns list of dicts with attention_pct, area_pct, attention_density.
|
|
"""
|
|
h, w = saliency.shape
|
|
total_saliency = saliency.sum()
|
|
total_area = h * w
|
|
|
|
if total_saliency == 0:
|
|
return [
|
|
{"attention_pct": 0.0, "area_pct": 0.0, "attention_density": 0.0}
|
|
for _ in regions
|
|
]
|
|
|
|
results = []
|
|
for region in regions:
|
|
rx, ry = region["x"], region["y"]
|
|
rw, rh = region["width"], region["height"]
|
|
|
|
# Clamp to image bounds
|
|
x1 = max(0, rx)
|
|
y1 = max(0, ry)
|
|
x2 = min(w, rx + rw)
|
|
y2 = min(h, ry + rh)
|
|
|
|
region_saliency = saliency[y1:y2, x1:x2].sum()
|
|
region_area = (x2 - x1) * (y2 - y1)
|
|
|
|
attention_pct = (region_saliency / total_saliency) * 100
|
|
area_pct = (region_area / total_area) * 100
|
|
density = attention_pct / max(area_pct, 0.01)
|
|
|
|
results.append({
|
|
"attention_pct": round(float(attention_pct), 2),
|
|
"area_pct": round(float(area_pct), 2),
|
|
"attention_density": round(float(density), 2),
|
|
})
|
|
|
|
return results
|