DATA COLLECTION CASE STUDY
Collection case study
From a scene gap.To a pilot plan.
Bring road types, lighting and subject coverage into one plan, then use pilot review to identify duplicates, exposure issues and coverage gaps.
All case studies
01 / PROJECT BRIEF
Agree on coverage before counting clips.
A road-perception team wants to add daytime, nighttime and suburban scenes. Start with targets, clip requirements and checks; use a pilot to refine the instructions before discussing production.
- Client profile
- road-perception research team
- Task
- Road-video collection plan
- Coverage
- Urban day, urban night and suburban day
- Current stage
- Collection planning · pilot review
02 / SCENE COVERAGE
Three scene groups. Distinct review priorities.
Keep the existing 60 / 30 / 30 clip allocation. Select a group to see its target, review priorities and the way a pilot record links to a scene.
URBAN · DAYTIME
Urban · daytime
Organize coverage around road users and junction types. Keep repeated scenes from occupying separate target slots.
- Target clips
- 60
- Target duration
- 1,200 sec
Pilot review focus
- Scene tags match junction and road-user requirements
- Clips decode, remain complete and meet the target duration
- Adjacent or duplicate content is identified separately
Review recordCOL-D-008 · A replacement passes recheck. The original retains its exclusion note and is not counted again.
URBAN · NIGHTTIME
Urban · nighttime
Include low-light usability and exposure in pilot checks. Define what needs recollection before increasing volume.
- Target clips
- 30
- Target duration
- 600 sec
Pilot review focus
- Visibility of intended subjects in low light
- Glare, reflections and exposure affecting the intended use
- Separate records for items requiring recollection
Review recordCOL-N-005 · An exposure issue awaits recollection review. It is excluded from the 23 currently passed records.
SUBURBAN · DAYTIME
Suburban · daytime
Track distant subjects and changing backgrounds. Reaching a clip count does not establish that coverage is complete.
- Target clips
- 30
- Target duration
- 600 sec
Pilot review focus
- Distant subjects meet agreed visibility criteria
- Variation in backgrounds and subject distance
- Scene tags link to their review evidence
Review recordCOL-S-004 · A replacement addresses distant-subject coverage. The recheck applies only to this individual record.
Each target clip lasts 20 seconds: 120 clips and 2,400 seconds across three groups. Coverage requirements and checks are tracked by scene.
03 / WORKING METHOD
Give every gap a clear next step.
Align scene and file requirements first. Pilot checks distinguish usable records from items needing more work; revisions remain linked to their original planning item.
- 01
Clip counts can conceal duplicates
ApproachGroup indexes by scene and identify repeated or highly similar content. Replacements keep the same planning item so old and new records are not counted twice.
Record retainedPlanning item, original record, duplicate reason and replacement reference
- 02
Nighttime usability is uncertain
ApproachUse pilot examples to agree on exposure and visibility. Keep items pending when evidence is insufficient and describe what needs to be supplied.
Record retainedCheck criteria, issue location, proposed action and recheck decision
- 03
Distant-subject coverage is hard to verify
ApproachInclude subject distance and background variation in the scene checklist, then link records to coverage items. Unsubstantiated coverage remains incomplete.
Record retainedScene tags, coverage evidence, gaps and instruction version
04 / PILOT REVIEW
Use pilot records to refine the next collection.
The preliminary review contains 24 records, split 12 / 6 / 6 across three scenes. Twenty pass initially; of four requiring action, three pass after replacement and one remains unresolved.
| Scene | Checks | Initial pass | Recheck pass / action items | Currently passed |
|---|---|---|---|---|
| Urban day | 12 | 11 | 1 / 1 | 12 |
| Urban night | 6 | 4 | 1 / 2 | 5 |
| Suburban day | 6 | 5 | 1 / 1 | 6 |
| Total | 24 | 20 | 3 / 4 | 23 |
Rates cover these 24 preliminary records only. Completion, coverage and acceptance of the subsequent 120-clip plan are recorded separately.
Explore delivery and quality05 / HANDOVER STRUCTURE
Connect the files, coverage and issue records.
Connect collection files, scene coverage and issue handling through batch indexes, linking each deliverable to its plan and review evidence.
- 01
Clips and batch indexes
Link each planning item to its clip, revision, duration and batch, keeping files and indexes aligned.
- 02
Scene and coverage manifest
Record road type, lighting, subject coverage and gaps. Distinguish target counts from verified counts.
- 03
Collection instructions and equipment references
Version the naming, encoding, resolution, equipment and timing requirements.
- 04
Review and recollection records
Retain initial checks, replacement references, rechecks and open issues without counting original and replacement records twice.
PROJECT TAKEAWAYS
Align three things before scaling a pilot.
- 01
Separate targets from completion
Use explicit definitions for planned clips, checks and actual completed collection.
- 02
Make replacements traceable
Link replacement records to the original issue and planning item, retaining exclusions to prevent double counting.
- 03
Keep coverage gaps visible
Passing some checks does not establish complete coverage. Retain recollection needs and unresolved items.
LET’S DEFINE YOUR COLLECTION TASK
Define the pilot around your missing scenes.
Describe the model task, scene gaps and expected volume to clarify what the first pilot needs to verify.