Adding Article on model_fcn_resnet

NEWS.md

@@ -15,6 +15,7 @@
 
 ## New features
 
+* Added article showcasing `model_fcn_resnet50()` with visualization utilities `draw_segmentation_masks()` and `make_vision_grid()` (@DerrickUnleashed, #244).
 * Added `imagenet_label()` and `imagenet_classes()` for ImageNet classes resolution (#229, @koshtiakanksha).
 * `base_loader()` now accept URLs (@cregouby, #246).
 

_pkgdown.yml

@@ -27,6 +27,8 @@ navbar:
         href: articles/examples/style-transfer.html
       - text: texture-nca
         href: articles/examples/texture-nca.html
+      - text: fcnresnet
+        href: articles/examples/fcnresnet.html
 
 reference:
 - title: Transforms

vignettes/examples/fcnresnet.R

@@ -0,0 +1,89 @@
+# Loading Images ---------------------------------------------------
+
+
+read_to_tensor <- function(url) {
+  tmp <- download_and_cache(url)
+  arr <- jpeg::readJPEG(tmp)              
+  arr <- as.array(arr)
+  torch_tensor(aperm(arr, c(3,1,2)), dtype = torch_float())  
+}
+
+url1 <- "https://raw.githubusercontent.com/pytorch/vision/main/gallery/assets/dog1.jpg"
+url2 <- "https://raw.githubusercontent.com/pytorch/vision/main/gallery/assets/dog2.jpg"
+
+dog1 <- read_to_tensor(url1)
+dog2 <- read_to_tensor(url2)
+
+
+# Visualizing a grid of images -------------------------------------
+
+
+dogs <- torch_stack(list(dog1, dog2))
+grid <- vision_make_grid(dogs, scale = TRUE, num_rows = 2)
+grid_arr <- as.array(grid$permute(c(2,3,1)))  
+plot(c(0, dim(grid_arr)[2]), c(0, dim(grid_arr)[1]), type = "n", ann = FALSE, axes = FALSE, asp = 1)
+rasterImage(grid_arr, 0, 0, w, h)
+
+
+# Preprocessing the data -------------------------------------
+
+
+norm_mean <- c(0.485, 0.456, 0.406)
+norm_std  <- c(0.229, 0.224, 0.225)
+
+preprocess <- function(img) {
+  resized <- nnf_interpolate(
+    img$unsqueeze(1), size = c(520, 520), mode = "bilinear", align_corners = FALSE
+  )$squeeze(1)
+  normed <- (resized - torch_tensor(norm_mean)$unsqueeze(2)$unsqueeze(3)) /
+              torch_tensor(norm_std)$unsqueeze(2)$unsqueeze(3)
+  list(resized = resized, normed = normed)
+}
+
+dog1_prep <- preprocess(dog1)
+dog2_prep <- preprocess(dog2)
+
+# make batch (2,3,520,520)
+input <- torch_stack(list(dog1_prep$normed, dog2_prep$normed))
+
+
+# Loading Model -------------------------------------
+
+
+model <- model_fcn_resnet50(pretrained = TRUE)
+model$eval()
+
+# run model
+output <- model(input)
+
+
+# Processing the Output ------------------------------
+
+
+# argmax over classes -> (1,H,W)
+mask_id <- output$out$argmax(dim = 2)
+
+make_masks <- function(mask_id_img, num_classes = 21) {
+  torch_stack(lapply(0:(num_classes-1), function(cls) mask_id_img$eq(cls)), dim = 1)
+}
+mask_bool1 <- make_masks(mask_id[1,..])
+mask_bool2 <- make_masks(mask_id[2,..])
+
+
+# Visualizing the Output ------------------------------
+
+
+segmented1 <- draw_segmentation_masks(
+  (dog1_prep$resized * 255)$to(torch_uint8()), 
+  masks = mask_bool1,
+  alpha = 0.6
+)
+
+segmented2 <- draw_segmentation_masks(
+  (dog2_prep$resized * 255)$to(torch_uint8()), 
+  masks = mask_bool2,
+  alpha = 0.6
+)
+
+tensor_image_browse(segmented1)
+tensor_image_browse(segmented2)

vignettes/examples/fcnresnet.Rmd

@@ -0,0 +1,9 @@
+---
+title: oxfordiiitpet-fcnresnet
+type: docs
+---
+
+```{r, echo = FALSE}
+knitr::opts_chunk$set(eval = FALSE)
+knitr::spin_child(paste0(rmarkdown::metadata$title, ".R"))
+```