Put the refinement on a slider, per layer

The evidence was being gathered and spent immediately at one strictness
nobody could see or change. This makes it a control.

`MaskLayer::refine` is shaping, like the feather, and it lives on the layer
for the layer's reason: two layers may sit on the same category and want
different amounts of it, and the model ran once for both.

## What the segmentation now stores

`CategorySummary` keeps the **coarse** mask and the `Refinement` beside it,
rather than a refined mask. That is what gives the control an off position
that is bit-for-bit the model's own weighting, and what stops a strictness
change from needing the model again.

`category_mask_at` borrows at zero and wherever no refinement could be
fitted, so a layer nobody has touched costs nothing over the old path.

The fit moved to the far side of the orientation permutation. The verdict is
a per-pixel field over the same grid as the mask it gates, so fitting it
upright would mean permuting a proxy-sized buffer afterwards to match — a
second rotation, and a second chance to get one wrong. One logit cell is the
same number of pixels either way: the letterbox scales by the longer edge and
a permutation does not change which edge that is.

The two frames became a `Frames` struct rather than six parameters. This
function reads the picture twice for opposite purposes — the model needs it
upright or it recognises far less, the refinement needs the sensor's grid —
and a transposed pair produces a plausible mask over slightly the wrong
pixels, which is the failure this module is most prone to.

## It rebuilds the field, and the signature says so

`refine` is mixed into `subject_signature`. Unlike a feather, which is read
off a field that is already correct, this changes which pixels are in the
mask at all — so it changes the coverage the field is measured from. Omitting
it is the bug where the slider moves and nothing happens until some unrelated
control invalidates the cache.

That puts it in the same cost class as a close or an open, which is why the
row takes `SliderRow::changed` — already once-per-gesture, since that row
takes `SliderTrack`'s `committed` internally — rather than a live stream.

The slider is offered only where there is something to move: a category
source, *and* a refinement the frame actually gave enough to fit. A control
that moves and does nothing is worse than an absent one.

## Two defaults that are deliberately different

A layer added from the panel starts at 4.0, because a category's edges are
twenty proxy pixels wide before anything is done to them and a photographer
adding a sky mask wants the sky rather than the sky plus every chimney in it.

A layer read from a sidecar with no `refine` key starts at **zero**. A file
written before this control existed has to render as it did then, and a
default of 4 on absence would quietly re-grade every stored category mask in
the catalogue. `a_categorys_refine_strictness_survives_and_defaults_off`
holds both halves, and `an_out_of_range_refine_is_clamped` holds the file to
the scale — past the top of it every colour fails and the mask deletes
itself, which reads as lost work rather than as a bad file.

`MAX_REFINE` is dr-pipeline's own constant mirroring
`dr_segment::STRICTNESS_MAX`, following `Falloff` and `Morphology`: this
crate holds the description of an edit and must not depend on the crate that
runs a model. dr-ui is where the two meet, and the only place that converts.

Verified: fmt clean, clippy --workspace -D warnings clean, 488 dr-pipeline
and 60 dr-segment tests.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

ui/dr-ui/src/segmentation.rs

@@ -30,6 +30,7 @@
 //! develop session, and [`crate::develop::SegmentationJob`] is the piece that
 //! carries the proxy render across with it.
 
+use std::borrow::Cow;
 use std::sync::Arc;
 
 use dr_gpu::GpuContext;
@@ -50,7 +51,21 @@ pub struct CategorySummary {
     pub coverage: f32,
     /// Coverage at proxy resolution, quantised to a byte — the same
     /// representation, and for the same reasons, as `InstanceSummary::mask`.
+    ///
+    /// **The model's own weighting, before any refinement.** Storing the
+    /// coarse mask rather than a refined one is what lets the refine control
+    /// have an off position that is exactly the old behaviour, and what stops
+    /// a strictness change from needing the model run again.
     pub mask: Vec<u8>,
+    /// The evidence for cutting this category back to the pixels whose colour
+    /// agrees with it, when there was enough of the frame to gather any.
+    ///
+    /// `None` is ordinary: a category thinner everywhere than the model's own
+    /// resolution, or one filling the whole frame, gives nothing to contrast
+    /// against — `dr_segment::refine` says which. A layer on such a category
+    /// simply has no refine control, which is the honest presentation of
+    /// "there was nothing to judge it with".
+    pub refinement: Option<dr_segment::Refinement>,
 }
 
 /// One recognised object.
@@ -130,6 +145,37 @@ impl Segmentation {
             .map(|c| c.mask.as_slice())
     }
 
+    /// One category's coverage, cut back at the given strictness.
+    ///
+    /// What the rasteriser reads. Borrowed at
+    /// [`dr_segment::STRICTNESS_OFF`] and wherever no refinement could be
+    /// fitted, so the common case — a layer whose control has not been moved —
+    /// costs nothing at all over [`Self::category_mask`].
+    ///
+    /// The strictness is the layer's, not the segmentation's: two layers may
+    /// sit on the same category and want different amounts of it, and the
+    /// model ran once for both.
+    pub fn category_mask_at(&self, name: &str, strictness: f32) -> Option<Cow<'_, [u8]>> {
+        let category = self.categories.iter().find(|c| &*c.name == name)?;
+        match &category.refinement {
+            Some(refinement) if strictness > dr_segment::STRICTNESS_OFF => Some(Cow::Owned(
+                refinement.apply_coverage(&category.mask, strictness),
+            )),
+            _ => Some(Cow::Borrowed(category.mask.as_slice())),
+        }
+    }
+
+    /// Whether this category has evidence a refine control could act on.
+    ///
+    /// The panel asks before offering the slider: a control that moves and
+    /// does nothing is worse than an absent one.
+    pub fn category_is_refinable(&self, name: &str) -> bool {
+        self.categories
+            .iter()
+            .find(|c| &*c.name == name)
+            .is_some_and(|c| c.refinement.is_some())
+    }
+
     /// Replace one instance in place, keeping every other index and the
     /// signature unchanged.
     ///
@@ -273,19 +319,23 @@ pub struct Options {
     pub fine: bool,
 
     /// TRACES: FR-DEV-3
-    /// Cut each scene category back to the pixels whose colour agrees with it.
+    /// Gather the evidence for cutting each scene category back to the pixels
+    /// whose colour agrees with it.
     ///
     /// The scene model's counterpart to [`Options::fine`], and it exists
     /// because tiling is not available here: a category has no bounding box to
     /// tile over — sky is wherever the sky is — so the only route to a sharper
     /// category edge is the photograph itself. See [`dr_segment::refine`].
     ///
+    /// This buys the *possibility* of refinement rather than any of it. What
+    /// it produces is a `dr_segment::Refinement` per category, which changes
+    /// no mask until a layer's refine control is moved off zero — so turning
+    /// this off costs the control, not the appearance of anything.
+    ///
     /// On by default where `fine` is off, because the two have opposite costs.
     /// Tiling is six inferences and a five-second wait; this is a distance
-    /// transform and a k-means over a subsample, tens of milliseconds against
-    /// a precompute already measured in hundreds. And it is subtractive, so
-    /// the worst it can do is take too much of a category rather than invent
-    /// one.
+    /// transform and a k-means over a subsample per category, tens of
+    /// milliseconds each against a precompute already measured in hundreds.
     pub refine: bool,
 }
 
@@ -370,7 +420,16 @@ pub fn compute(
     // way. Failures here are logged and dropped rather than propagated: no
     // scene model is an ordinary state, and a photograph that can be masked by
     // subject should not become unopenable because the categories are absent.
-    let categories = match scene_categories(&stood_up, uw, uh, orientation, options.refine) {
+    let categories = match scene_categories(
+        &Frames {
+            upright: &stood_up,
+            upright_size: (uw, uh),
+            sensor: rgb,
+            sensor_size: (width, height),
+        },
+        orientation,
+        options.refine,
+    ) {
         Ok(c) => c,
         Err(e) => {
             log::info!("no scene categories for this frame: {e}");
@@ -545,6 +604,27 @@ fn detect(
         .map_err(|e| e.to_string())
 }
 
+/// The one photograph, in the two frames this pass needs it in.
+///
+/// Both, and not one plus a permutation applied where needed, because this
+/// function reads the picture *twice* for different purposes and the two want
+/// opposite frames. The model must be shown an upright photograph or it
+/// recognises far less (see [`upright`]); the refinement must read the sensor's
+/// own grid, because the field it produces has to line up pixel for pixel with
+/// a mask that has already been laid back down.
+///
+/// Carried as a struct rather than six parameters so that a caller cannot
+/// quietly transpose the pair — which would produce a plausible mask over
+/// slightly the wrong pixels, the failure this whole module is most prone to.
+struct Frames<'a> {
+    /// As the photographer sees it. What the model reads.
+    upright: &'a [f32],
+    upright_size: (usize, usize),
+    /// As the sensor wrote it. The space every stored mask lives in.
+    sensor: &'a [f32],
+    sensor_size: (usize, usize),
+}
+
 /// Weigh the photographic categories, if this build can find a scene model.
 ///
 /// Separate from [`detect`] rather than folded into it because the two are
@@ -552,15 +632,16 @@ fn detect(
 /// frame with no recognisable subject still has sky. Neither failure should
 /// take the other down.
 fn scene_categories(
-    rgb: &[f32],
-    width: usize,
-    height: usize,
+    frames: &Frames<'_>,
     orientation: Orientation,
     refine: bool,
 ) -> Result<Vec<CategorySummary>, String> {
+    let (upright, (width, height)) = (frames.upright, frames.upright_size);
+    let (sensor, (sensor_width, sensor_height)) = (frames.sensor, frames.sensor_size);
+
     let mut model = load_scene_model()?;
     let scene = model
-        .analyse(rgb, width, height)
+        .analyse(upright, width, height)
         .map_err(|e| e.to_string())?;
 
     let mut out = Vec::new();
@@ -577,41 +658,6 @@ fn scene_categories(
             continue;
         };
 
-        // Cut the category back to the pixels whose colour agrees with it.
-        //
-        // Here rather than after `lay_down` because this reads the photograph,
-        // and the photograph is upright at this point — refining on the far
-        // side of the permutation would mean carrying a second, rotated copy
-        // of the proxy across for it to read.
-        let weights = if refine {
-            let (refined, what) = dr_segment::refine_category(
-                &weights,
-                rgb,
-                width,
-                height,
-                scene.cell_pixels(),
-                &dr_segment::RefineOptions::default(),
-            );
-            match what {
-                dr_segment::Refined::Applied { removed } => {
-                    log::debug!(
-                        "refined '{name}': cut {:.1}% of its weight",
-                        removed * 100.0
-                    );
-                }
-                // An ordinary state, not a failure — see `dr_segment::refine`.
-                // Logged all the same, because a refinement that silently did
-                // nothing is indistinguishable from the feature being off, and
-                // "the sky looks the way it did before" is what both look like.
-                dr_segment::Refined::Skipped(why) => {
-                    log::debug!("left '{name}' coarse: {why:?}");
-                }
-            }
-            refined
-        } else {
-            weights
-        };
-
         // Back into sensor space, exactly as an instance mask is: the grid a
         // stored layer indexes into has to be the sensor's whatever the model
         // was shown. `lay_down` wants a box too, so it gets the whole frame —
@@ -623,10 +669,47 @@ fn scene_categories(
             height,
             orientation,
         );
+
+        // Gather the evidence for cutting this category back, on the far side
+        // of the permutation.
+        //
+        // Sensor space rather than the model's, and that is the whole reason
+        // this happens here: the verdict is a per-pixel field over the same
+        // grid as the mask it gates, so fitting it upright would mean
+        // permuting it afterwards to match — a second rotation of a
+        // proxy-sized buffer, and a second chance to get one wrong.
+        //
+        // One logit cell is the same number of pixels in either frame: the
+        // letterbox scales by the *longer* edge, and a permutation does not
+        // change which edge that is.
+        let refinement = refine
+            .then(|| {
+                dr_segment::Refinement::compute(
+                    &mask,
+                    sensor,
+                    sensor_width,
+                    sensor_height,
+                    scene.cell_pixels(),
+                    &dr_segment::RefineOptions::default(),
+                )
+            })
+            .and_then(|result| match result {
+                Ok(refinement) => Some(refinement),
+                // An ordinary state, not a failure — see `dr_segment::refine`.
+                // Logged all the same, because the layer will simply have no
+                // refine control and "there is no slider" should be traceable
+                // to a reason rather than looking like a missing feature.
+                Err(why) => {
+                    log::debug!("no refinement for '{name}': {why:?}");
+                    None
+                }
+            });
+
         out.push(CategorySummary {
             name: name.clone(),
             coverage,
             mask: quantise(&mask),
+            refinement,
         });
     }