05. GUI application for visualizing (adder viz)
Follow the instructions here
Simply drag an .mp4, .aedat4, or .dat file into the Transcode window to begin live transcoding.
- Δt_ref: ticks per reference interval. Should be 255 for an 8-bit framed source, such that each frame is said to span 255 ticks of time.
- Δt_max multiplier: Δt_max := Δt_ref * [this number]. The maximum number of ticks that the first event at a new intesnity level may span.
- ADΔER threshold: the number of intensity units by which an incoming intensity may vary from the pixel's baseline before events must be fired. If threshold is 0, pixel will fire its accumulated events as soon as the incoming intensity changes. If threshold is high, then incoming intensities are averaged out over time into fewer ADΔER events.
- Thread count: How many hardware threads to use.
- Video scale: For framed video, what scaling factor to use to shrink images before integrating.
- Channels: For framed video, use color or black-and-white?
- Integration mode: Normal = try to represent the incoming intensity variations/timing as accurately as possible (slower); Collapse = Only build up and output 1 informative event each time the intensity changes across the threshold.
- View mode: In the transcoder window, which view to display. (Normalized D components, normalized Δt components, or intensity = 2^D/Δt)
- Compression mode: What to do with the events that are generated. Empty = throw them into the void; Raw = write them out directly to a binary file; Compressed = perform source-modeled compression as in this paper.
- Event output order: Optionally temporally interleave the events (so that multiple back-to-back events from any pixel are interleaved with the events from other pixels)
- Detect features: Perform source-agnostic feature detection. Enabling this makes the pixels close to a newly detected feature more sensitive. This radius is determined by the feature radius slider or CRF quality. Below this checkbox, you can choose if/how to visualize the features.
- DAVIS mode: If source is a DVS/DAVIS video, which method to use for transcoding.
- DAVIS deblurred FPS: Sets the timespan to use for deblurred DAVIS APS images. If 1000, for example, then each deblurred frame will span 1000 ticks with one second of the video spanning 1E6 ticks.
- Optimize: If source is a DAVIS video, optimize the contrast threshold parameter?
- Metrics: For framed video sources, optional frame-based quality metrics to compared the transcoded representation to the input video.
Click the Play file tab on the menu bar, then drag and drop your .adder file into the window. Playback will begin immediately. You can adjust the playback speed and the image view dynamically. The frame buffer caps the latency by a certain number of reconstructed frames, which is useful if you're playing back video with extremely sparse events.