dtourolle
0da8271836
Local masking needs to know where an image's regions are. The watershed spike (S15 arm A) found the boundaries but had no idea what any of them enclosed; its coarse levels were geometric accidents. This adds the other half and the thing that joins them. `core/dr-segment` is where region reasoning now lives — the hierarchy moves out of `dr-gpu`, which keeps only the pixel passes that are genuinely shaders. The new crate is device-free and, without its default features, model-free too: 20 of its tests need neither an adapter nor 11 MB of weights. Arm B runs YOLO26n-seg through `ort`. D13 framed inference as a choice between `ort`'s C++ runtime and the pure-Rust dependency policy; that was a false choice. `ort`'s `alternative-backend` feature unlinks the C entirely and `ort-tract` supplies the API from tract, which is pure Rust. Measured before committing to it: zero unsupported operators, 420 ms for 640x640, and correct masks on bus.jpg. No NDK problem to solve, so D13's largest tolerated exception is not needed. Arm C is `prior.rs`, and it ships because the two arms fail in opposite directions. Instance membership re-weights the merge saddles, so region pairs the model believes share an object merge early and pairs straddling its edge merge late. No boundary moves — only the order in which they dissolve — which is how the result stays pixel-accurate at every level while its coarse levels become named things. Two things the spec assumed that turned out to be false, both recorded in models/LICENCE.md: there is no usable ADE20K-trained YOLO, so the shipped vocabulary is COCO's 80 subjects and *stuff* like sky and foliage must come from arm A; and tract cannot parse a dynamic-shape export, so the graph's input is fixed and tiling is the only route to more semantic resolution. Weights are AGPL-3.0, which GPLv3 §13 permits and which makes the combined work effectively AGPL. Deliberate, not accidental. They live in Git LFS, and a build script fails with an instruction rather than embedding a pointer file when the clone lacks them.
47 lines
1.3 KiB
TOML
47 lines
1.3 KiB
TOML
[package]
|
|
name = "dr-gpu"
|
|
version.workspace = true
|
|
edition.workspace = true
|
|
rust-version.workspace = true
|
|
license.workspace = true
|
|
|
|
[dependencies]
|
|
dr-types.workspace = true
|
|
dr-decode.workspace = true
|
|
dr-pipeline.workspace = true
|
|
# The watershed's pixel passes are here because they are shaders; everything
|
|
# that reasons about regions rather than pixels lives there, where it is
|
|
# testable with no adapter present. No features: this half needs neither the
|
|
# inference runtime nor the weights, and the workspace declaration defaults
|
|
# them off so that stays true.
|
|
dr-segment.workspace = true
|
|
wgpu.workspace = true
|
|
thiserror.workspace = true
|
|
log.workspace = true
|
|
bytemuck.workspace = true
|
|
# Needed outside tests: shader compilation errors are collected through an
|
|
# async error scope, which must be resolved before the pipeline is returned.
|
|
pollster.workspace = true
|
|
|
|
[dev-dependencies]
|
|
env_logger.workspace = true
|
|
|
|
[[example]]
|
|
name = "bench"
|
|
required-features = ["readback"]
|
|
|
|
[[example]]
|
|
name = "develop"
|
|
# No `readback` needed since S1: this writes a file, so it goes through
|
|
# `export_pixels`, which is ungated precisely because an export is not the
|
|
# round-trip AC-8 forbids.
|
|
|
|
[features]
|
|
default = []
|
|
# Exposes read_pixels outside tests. Production must not enable this.
|
|
readback = []
|
|
|
|
[[example]]
|
|
name = "segment"
|
|
required-features = ["readback"]
|