/dynafit
Personal Research
DYNAFIT
Exploring how smartphone video could support fit decisions for a body in motion.
/CONTRIBUTION
I developed the research direction and iterated the pipeline architecture, including a human-review boundary.
ROLE
Personal research and pipeline architecture exploration
STATUS
Ongoing research; end-to-end outputs unestablished
/dynafit
Personal Research
DYNAFIT
Exploring how smartphone video could support fit decisions for a body in motion.
/CONTRIBUTION
I developed the research direction and iterated the pipeline architecture, including a human-review boundary.
ROLE
Personal research and pipeline architecture exploration
STATUS
Ongoing research; end-to-end outputs unestablished
/dynafit
Personal Research
DYNAFIT
Exploring how smartphone video could support fit decisions for a body in motion.
/CONTRIBUTION
I developed the research direction and iterated the pipeline architecture, including a human-review boundary.
ROLE
Personal research and pipeline architecture exploration
STATUS
Ongoing research; end-to-end outputs unestablished
STATIC
DINED static Persona 1 / not a Dynafit video or pipeline output.
STATIC
DINED static Persona 1 / not a Dynafit video or pipeline output.
01
RESEARCH QUESTION
Study the body in motion
A plausible reconstruction is not yet biomechanical evidence.
01
RESEARCH QUESTION
Study the body in motion
A plausible reconstruction is not yet biomechanical evidence.
Designing for bodies as they move
Static body references can miss geometry changing during movement. Dynafit explores whether smartphone video could help designers inspect movement-related fit questions; no accessibility or accuracy advantage is evaluated.
Separate reconstruction from validation
RATIONALE
Human inspection should sit between visual reconstruction and biomechanical interpretation. Plausible geometry alone is not valid measurement.
CONSEQUENCE
The current evidence is a framework account, not a validated sensing or clinical system.
Designing for bodies as they move
Static body references can miss geometry changing during movement. Dynafit explores whether smartphone video could help designers inspect movement-related fit questions; no accessibility or accuracy advantage is evaluated.
Separate reconstruction from validation
RATIONALE
Human inspection should sit between visual reconstruction and biomechanical interpretation. Plausible geometry alone is not valid measurement.
CONSEQUENCE
The current evidence is a framework account, not a validated sensing or clinical system.
02
PROPOSED PIPELINE
From video to a reviewed model
Five proposed stages, with human review kept explicit.
02
PROPOSED PIPELINE
From video to a reviewed model
Five proposed stages, with human review kept explicit.
The source describes four iterations and identifies Pipeline 2.0 as the fourth. Records for iterations one to three and comparison criteria are unavailable. Its reported architectural shift separates visual reconstruction from biomechanical validity and introduces human review; this is not a validated run.
The source describes four iterations and identifies Pipeline 2.0 as the fourth. Records for iterations one to three and comparison criteria are unavailable. Its reported architectural shift separates visual reconstruction from biomechanical validity and introduces human review; this is not a validated run.
Before review → Capture and reconstruct
Source-named pre-review stages. Capture, calibration and integration runs are unknown.
- 01→
Capture and calibrate
Smartphone video, focus lock, ChArUco and undistortion are named; actual capture and calibration runs are unknown.
PROPOSED
- 02[ END ]
Reconstruct
SAM 3D Body / MHR are named directions; integration and a resulting reconstruction are unestablished.
PROPOSED
/REVIEW BOUNDARY
03 / Human review →
RATIONALE
Inspect and correct the reconstructed body before interpretation. This is an architectural direction.
CONSEQUENCE
The correction method and implementation are unknown; no corrected model or recorded handoff is supplied. A plausible reconstruction is not biomechanical evidence.
After review → Interpret and derive questions
Source-named post-review directions; task validation and derived outputs remain unestablished.
- 01→
Biomechanical interpretation
SKEL / SUPR, inverse kinematics and scene relationships are named; task validation is unestablished.
PROPOSED
- 02[ END ]
Intended design outputs
Temporal geometry, kinematic representation, contact interpretation and reports are intended, not supplied artifacts.
INTENDED
Before review → Capture and reconstruct
Source-named pre-review stages. Capture, calibration and integration runs are unknown.
- 01→
Capture and calibrate
Smartphone video, focus lock, ChArUco and undistortion are named; actual capture and calibration runs are unknown.
PROPOSED
- 02[ END ]
Reconstruct
SAM 3D Body / MHR are named directions; integration and a resulting reconstruction are unestablished.
PROPOSED
Keep correction inspectable
RATIONALE
Human review is the architectural boundary between reconstruction and downstream interpretation.
Define reference and error checks
A task-specific reference, explicit error measures and inspection records are needed before a reconstructed body can support a design decision.
Keep correction inspectable
RATIONALE
Human review is the architectural boundary between reconstruction and downstream interpretation.
Define reference and error checks
A task-specific reference, explicit error measures and inspection records are needed before a reconstructed body can support a design decision.
03
REFERENCE
Start with a body reference
An existing body reference motivates the question.
03
REFERENCE
Start with a body reference
An existing body reference motivates the question.
A static DINED Persona 1 reference motivates the dynamic-body question but says nothing about Dynafit output quality.
A static DINED Persona 1 reference motivates the dynamic-body question but says nothing about Dynafit output quality.
04
METHODS
Methods and planned outputs
Named methods and intended outputs remain distinct from demonstrated integration.
04
METHODS
Methods and planned outputs
Named methods and intended outputs remain distinct from demonstrated integration.
ChArUco / undistortion — Proposed capture preparation. Status: RUN UNKNOWN. The input clip, executable run, resulting output and reference comparison are absent. These methods do not establish accuracy, runtime, deployment or pressure results.
ChArUco / undistortion — Proposed capture preparation. Status: RUN UNKNOWN. The input clip, executable run, resulting output and reference comparison are absent. These methods do not establish accuracy, runtime, deployment or pressure results.
ChArUco / undistortion — Proposed capture preparation. Status: RUN UNKNOWN. The input clip, executable run, resulting output and reference comparison are absent. These methods do not establish accuracy, runtime, deployment or pressure results.
Intended outputs
Proposed application links: facial surface change during speech → mask-fit exploration; foot movement through gait → footwear-fit exploration. Neither use nor a pressure result is evaluated or validated.
Mesh / speech → mask fit
Temporal surface geometry
Proposed question: how does facial surface change during speech inform mask fit? No output or fit evaluation is supplied.
Joints / gait → footwear fit
Kinematic representation
Proposed question: how does foot form change through gait inform footwear fit? No integrated output or validated use is supplied.
Contact / pressure
Intended interpretation
No measured pressure output or validated result is supplied.
Report
Inspectability and evidence
Intended; no completed pipeline report is supplied.
Intended outputs
Proposed application links: facial surface change during speech → mask-fit exploration; foot movement through gait → footwear-fit exploration. Neither use nor a pressure result is evaluated or validated.
Mesh / speech → mask fit
Temporal surface geometry
Proposed question: how does facial surface change during speech inform mask fit? No output or fit evaluation is supplied.
Joints / gait → footwear fit
Kinematic representation
Proposed question: how does foot form change through gait inform footwear fit? No integrated output or validated use is supplied.
Contact / pressure
Intended interpretation
No measured pressure output or validated result is supplied.
Report
Inspectability and evidence
Intended; no completed pipeline report is supplied.
Intended outputs
Proposed application links: facial surface change during speech → mask-fit exploration; foot movement through gait → footwear-fit exploration. Neither use nor a pressure result is evaluated or validated.
Mesh / speech → mask fit
Temporal surface geometry
Proposed question: how does facial surface change during speech inform mask fit? No output or fit evaluation is supplied.
Joints / gait → footwear fit
Kinematic representation
Proposed question: how does foot form change through gait inform footwear fit? No integrated output or validated use is supplied.
Contact / pressure
Intended interpretation
No measured pressure output or validated result is supplied.
Report
Inspectability and evidence
Intended; no completed pipeline report is supplied.
05
NEXT STEPS
What a complete run must show
Record a complete run before interpreting performance metrics.
05
NEXT STEPS
What a complete run must show
Record a complete run before interpreting performance metrics.
An architecture with explicit evidence boundaries
DIRECTION
Reconstruct, review, interpret
The framework keeps capture, reconstruction, human inspection and downstream interpretation distinct.
LIMIT
No demonstrated end-to-end result
Actual pipeline outputs, accuracy, runtime and deployment are not established.
NEXT
Record one complete chain
Use an actual input clip, record the run and compare the output against a task-relevant reference before interpreting derived metrics.
An architecture with explicit evidence boundaries
DIRECTION
Reconstruct, review, interpret
The framework keeps capture, reconstruction, human inspection and downstream interpretation distinct.
LIMIT
No demonstrated end-to-end result
Actual pipeline outputs, accuracy, runtime and deployment are not established.
NEXT
Record one complete chain
Use an actual input clip, record the run and compare the output against a task-relevant reference before interpreting derived metrics.
An architecture with explicit evidence boundaries
DIRECTION
Reconstruct, review, interpret
The framework keeps capture, reconstruction, human inspection and downstream interpretation distinct.
LIMIT
No demonstrated end-to-end result
Actual pipeline outputs, accuracy, runtime and deployment are not established.
NEXT
Record one complete chain
Use an actual input clip, record the run and compare the output against a task-relevant reference before interpreting derived metrics.