Multi-Axis Quality Scoring Framework

Discover how the BestShots multi-axis scoring framework evaluates technical quality, balances genre criteria, and maps images to 5-star ratings.

Technical Evaluation Dimensions

BestShots evaluates photographs using an advanced computer vision engine tuned specifically for professional photography standards. Rather than relying on a single generic score, each frame is analyzed across 8 distinct criteria scored on a 1 to 5 scale (1 = poor, 3 = acceptable, 5 = excellent):

Dimension	Display Label	Baseline Weight	Key Evaluation Heuristics
Subject Focus	Subject Focus	0.25	Eyelash and iris sharpness, focal plane alignment, depth-of-field accuracy.
Expression & Moment	Expression & Moment	0.15	Eye openness, authentic smiles, absence of awkward blinks, decisive action timing.
Dynamic Range & Exposure	Dynamic Range & Exposure	0.15	Highlight detail preservation, shadow recoverability, clipping prevention.
Composition & Framing	Composition & Framing	0.10	Rule of thirds, leading lines, horizon level, headroom, edge cleanliness.
Color Fidelity & Skin Tones	Color Fidelity & Skin Tones	0.10	Natural skin tone rendition under mixed ambient and strobe lighting.
Motion Control & Freeze	Motion Control & Freeze	0.10	Controlled action freezing versus unintended camera shake or subject blur.
Noise & Grain Control	Noise & Grain Control	0.10	High-ISO grain management, color banding, and compression artifacts.
Clean Background	Clean Background	0.05	Lens flare, sensor dust spots, chromatic aberration, background clutter.

Dynamic Shoot Profile Calibration

While the table above reflects the balanced general culling profile (Other), BestShots dynamically calibrates these weights based on your selected shoot profile.

For example, when evaluating a Sports shoot, motion freeze and subject tracking take priority over background cleanliness:

Primary Culling Focus: Peak action freezing, motion control, and decisive moments.
Curation Strategy: Athletic timing, subject tracking sharpness, and high-speed motion freezing.
Calibrated Weight Distribution:
- Focus & Subject Sharpness: 20%
- Motion Control & Freeze: 25%
- Subject Expression & Moment: 15%
- Composition & Framing: 15%
- Exposure & Dynamic Range: 10%
- Color & Skin Tones: 5%
- Noise & Grain Control: 5%
- Clean Background: 5%

Note

Because weights strictly sum to 1.00 in all 9 shoot profiles, score calculation remains mathematically normalized regardless of which genre is selected. See Shoot Profile Calibrations for the complete weight matrix.

Normalized Scoring & Decision Mapping

The AI engine computes an overall weighted quality score for each image. Each of the 8 criteria is scored from 1 to 5, multiplied by its profile weight, and scaled to produce a normalized score between 20 and 100.

Strict Decision Mapping

The calculated score directly maps to an initial 5-star rating and culling recommendation:

Score Range	Star Rating	Decision	Evaluation Criteria
0 to 34	1 Star	REJECT	Unrecoverable issues: subject out of focus, severe camera blur, blown highlights on faces, eyes closed across key subjects, or clearly inferior duplicate.
35 to 49	2 Stars	REJECT	Probably discard: usable only for minimal coverage needs, with notable post-production fixes required.
50 to 69	3 Stars	KEEP	Baseline keeper: technically acceptable, but requires review before final delivery.
70 to 84	4 Stars	SELECT	Strong candidate: solid technical execution, sharp subject, and compelling moment.
85 to 100	5 Stars	HERO	Standout image: portfolio-ready, exceptional sharpness, and decisive composition.

Important

Initial AI decisions are clear and decisive (Reject, Keep, Select, or Hero). Ambiguous states like "maybe" are avoided so you can filter your catalog instantly.

Multi-Stage Verification & Quality Integrity

To ensure objective and repeatable culling accuracy, the AI engine performs a structured multi-stage quality review before presenting ratings:

Detailed Visual Inspection: The system inspects optical sharpness, lighting, facial expressions, and composition in full detail before assigning individual category marks.
Four-Stage Verification Checkpoints:
- Stage 1 (Subject Identification): Confirms primary subject identification and applies the shoot profile calibration.
- Stage 2 (Evidence Validation): Verifies observable optical evidence for each score, confirming weights sum to 1.00.
- Stage 3 (Consistency Check): Enforces that the overall score aligns with rating and decision thresholds.
- Stage 4 (Metadata Integrity): Validates all structured fields before ratings appear in your desktop workstation.

Studio Privacy & Security Guarantee

All image evaluations are executed through private, secure cloud infrastructure:

Strict Zero AI Training Guarantee: Client images operate under strict zero-data-retention policies. Your client images are never used to train or fine-tune public AI models.
High-Fidelity Optical Inspection: Native proxy processing ensures fine details like iris focus and fabric texture are accurately evaluated without compression distortion.
Enterprise Isolation: Private dedicated processing ensures fast turnarounds during heavy peak-season upload volumes.