In the computational cluster we had Nanne van Noord, Carlo Bretti, and Andrea Farina. They shared with the workshop the challenges of analysing moving images using units at different scales, e.g. pixels, shots, scenes, films, archives.See for example this table from my article on Creanalytics:
| Artefact | Processing level | Example |
|---|---|---|
| Cinema | Social – aggregate | Popular Hollywood cinema |
| Film | Human | Jurassic Park (1993) |
| Clip | Human/computer | Raptors in the Kitchen Scene (https://youtu.be/dnRxQ3dcaQk) |
| Shot | Human-computer | 130 frames (5.421 s) |
| Frame | Computer/human | Individual frame (512 × 340 pixels) |
| Pixels | Numeric – disaggregate | Vector ([176800x1]); Tensor ([16, 3, 340, 512]) |
This cluster also highlighted the difficulties of encoding weak conventions or tacit knowledge, which echoes some of the concerns of the collections cluster and the work of Pauline. Later, during the ideation sessions, we came back for example to the idea of detecting “scenes” or “blocs” which are defined by semantic convention and defined annotation guidelines for this kind of unit.