Update interpolate

mistralrs-core/src/vision_models/phi3_inputs_processor.rs

@@ -5,7 +5,7 @@ use std::{any::Any, sync::Arc};
 use candle_core::{Device, Result, Tensor};
 use image::{imageops::FilterType, DynamicImage, GenericImage, GenericImageView, Rgba};
 use itertools::Itertools;
-use mistralrs_vision::{ApplyTransforms, Normalize, ToTensor, ToTensorAndResize, Transforms};
+use mistralrs_vision::{ApplyTransforms, InterpolateResize, Normalize, ToTensor, Transforms};
 use regex_automata::meta::Regex;
 use tokenizers::Tokenizer;
 
@@ -390,15 +390,17 @@ impl ImagePreProcessor for Phi3InputsProcessor {
             };
             // Transforms for the global image (after HD transform, resized)
             let transforms_global = Transforms {
-                input: &ToTensorAndResize {
-                    target_h: 336,
-                    target_w: 336,
-                    filter: FilterType::CatmullRom,
-                },
-                inner_transforms: &[&Normalize {
-                    mean: config.image_mean.unwrap_or(Self::DEFAULT_MEAN).to_vec(),
-                    std: config.image_std.unwrap_or(Self::DEFAULT_STD).to_vec(),
-                }],
+                input: &ToTensor,
+                inner_transforms: &[
+                    &Normalize {
+                        mean: config.image_mean.unwrap_or(Self::DEFAULT_MEAN).to_vec(),
+                        std: config.image_std.unwrap_or(Self::DEFAULT_STD).to_vec(),
+                    },
+                    &InterpolateResize {
+                        target_h: 336,
+                        target_w: 336,
+                    },
+                ],
             };
 
             // Resize with bicubic interpolation

mistralrs-vision/src/lib.rs

@@ -2,7 +2,7 @@ use candle_core::{Device, Result, Tensor};
 use image::DynamicImage;
 mod transforms;
 pub(crate) mod utils;
-pub use transforms::{Normalize, ToTensor, ToTensorAndResize};
+pub use transforms::{InterpolateResize, Normalize, ToTensor};
 
 pub trait ImageTransform {
     type Input;

mistralrs-vision/src/transforms.rs

@@ -1,9 +1,6 @@
 use crate::utils::{get_pixel_data, n_channels};
 use candle_core::{DType, Device, Result, Tensor};
-use image::{
-    imageops::{self, FilterType},
-    DynamicImage, GenericImageView,
-};
+use image::{DynamicImage, GenericImageView};
 
 use crate::ImageTransform;
 
@@ -77,21 +74,17 @@ impl ImageTransform for Normalize {
 
 /// Do what `ToTensor` does, but also resize the image without preserving
 /// aspect ratio.
-pub struct ToTensorAndResize {
+pub struct InterpolateResize {
     pub target_w: usize,
     pub target_h: usize,
-    pub filter: FilterType,
 }
 
-impl ImageTransform for ToTensorAndResize {
-    type Input = DynamicImage;
-    type Output = Tensor;
+impl ImageTransform for InterpolateResize {
+    type Input = Tensor;
+    type Output = Self::Input;
 
-    fn map(&self, x: &Self::Input, device: &Device) -> Result<Self::Output> {
-        let n_channels = n_channels(x);
-        let img = imageops::resize(x, self.target_w as u32, self.target_h as u32, self.filter);
-        let data = get_pixel_data(n_channels, img, self.target_h, self.target_w);
-        ToTensor::to_tensor(device, n_channels, data)
+    fn map(&self, x: &Self::Input, _: &Device) -> Result<Self::Output> {
+        x.interpolate2d(self.target_h, self.target_w)
     }
 }