fix: undo additional error checking. this will be pushed into the iph…

…one app instead.

src/fish/fish_segmentation.rs

@@ -10,7 +10,7 @@ use ndarray::{s, Array2, Array3, ArrayBase, Axis, Dim, IxDynImpl, OwnedRepr};
 use opencv::core::{Mat, Point2i, Size, VectorToVec, CV_8UC1};
 use opencv::imgproc::{fill_poly, find_contours_with_hierarchy, resize_def, CHAIN_APPROX_NONE, LINE_8, RETR_CCOMP};
 use opencv::types::{VectorOfPoint, VectorOfVec4i, VectorOfVectorOfPoint};
-use ort::{GraphOptimizationLevel, Session, SessionOutputs};
+use ort::{GraphOptimizationLevel, Session};
 use reqwest::blocking::get;
 use cv_convert::TryIntoCv;
 
@@ -275,8 +275,17 @@ impl FishSegmentation {
  }
  }
 
- fn parse_results_from_inference(&self, outputs: SessionOutputs) ->
+ fn do_inference(&self, img: &Array3<f32>, model: &Session) ->
  Result<(ArrayBase<OwnedRepr<f32>, Dim<IxDynImpl>>, ArrayBase<OwnedRepr<f32>, Dim<IxDynImpl>>, ArrayBase<OwnedRepr<f32>, Dim<IxDynImpl>>), ort::Error> {
+ let mut clone = img.clone();
+ clone.swap_axes(2, 1);
+ clone.swap_axes(1, 0);
+
+ println!("RUST: clone: {}, {}, {}", clone.shape()[0], clone.shape()[1], clone.shape()[2]);
+
+ println!("RUST: Before Run");
+ let outputs = model.run(ort::inputs!["argument_1.1" => clone.view()]?)?;
+ println!("RUST: After Run");
 
  // boxes=tensor18, classes=pred_classes, masks=5232, scores=2339, img_size=onnx::Split_174
  println!("RUST: Before parsing results");
@@ -292,45 +301,6 @@ impl FishSegmentation {
  Ok((boxes, masks, scores))
  }
 
- fn do_inference(&self, img: &Array3<f32>, model: &Session) ->
- Result<(ArrayBase<OwnedRepr<f32>, Dim<IxDynImpl>>, ArrayBase<OwnedRepr<f32>, Dim<IxDynImpl>>, ArrayBase<OwnedRepr<f32>, Dim<IxDynImpl>>), SegmentationError> {
- let mut clone = img.clone();
- clone.swap_axes(2, 1);
- clone.swap_axes(1, 0);
-
- println!("RUST: clone: {}, {}, {}", clone.shape()[0], clone.shape()[1], clone.shape()[2]);
-
- println!("RUST: Before Run");
- match ort::inputs!["argument_1.1" => clone.view()] {
- Ok(input) => {
- match model.run(input) {
- Ok(outputs) => {
- println!("RUST: After Run");
-
- match self.parse_results_from_inference(outputs) {
- Ok((boxes, masks, scores)) => Ok((boxes, masks, scores)),
- Err(err) => {
- match err {
- ort::Error::PointerShouldBeNull(err) => {
- if err == "GetTensorMutableData" {
- Err(SegmentationError::FishNotFound)
- }
- else {
- Err(SegmentationError::OrtErr(ort::Error::PointerShouldBeNull(err)))
- }
- },
- other => Err(SegmentationError::OrtErr(other))
- }
- }
- }
- },
- Err(err) => Err(SegmentationError::OrtErr(err))
- }
- },
- Err(err) => Err(SegmentationError::OrtErr(err))
- }
- }
-
  fn do_paste_mask(&self, masks: &Array2<f32>, img_h: u32, img_w: u32) -> Result<Array2<f32>, SegmentationError> {
  let masks_unsqueezed = masks.clone().insert_axis(Axis(2));
 
@@ -513,14 +483,7 @@ impl FishSegmentation {
 
  Ok(masks)
  }
- Err(error) => {
- match error {
- SegmentationError::FishNotFound => {
- Ok(Array2::<u8>::zeros((orig_height, orig_width)))
- },
- other => Err(other)
- }
- }
+ Err(error) => Err(SegmentationError::OrtErr(error))
  }
  }
 }