Add files via upload

Author: WevertonGomesCosta

pig_weight_calculation_by_mask_r_cnn_keras_and_tensorflow.py

@@ -0,0 +1,322 @@
+# -*- coding: utf-8 -*-
+"""Pig-weight-calculation-by-Mask-R-CNN-Keras-and-TensorFlow.ipynb
+
+Automatically generated by Colaboratory.
+
+Original file is located at
+    https://colab.research.google.com/drive/1SsI1zEu7hiTZi3YbUdA61Sqf1ZqXz-eF
+
+## **1. Installation**
+
+Git clone repositorie and install libraries
+"""
+
+# Update CUDA for TF 2.5
+!wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/libcudnn8_8.1.0.77-1+cuda11.2_amd64.deb
+!dpkg -i libcudnn8_8.1.0.77-1+cuda11.2_amd64.deb
+# Check if package has been installed
+!ls -l /usr/lib/x86_64-linux-gnu/libcudnn.so.*
+# Upgrade Tensorflow
+!pip install --upgrade tensorflow==2.5.0
+
+# Commented out IPython magic to ensure Python compatibility.
+!git clone https://github.com/WevertonGomesCosta/Pig-weight-calculation-by-Mask-R-CNN-Keras-and-TensorFlow.git
+# %matplotlib inline
+import warnings
+warnings.filterwarnings('ignore')
+
+"""## **2. Root path and functions**
+Declare root path and import functions on m_rcnn
+"""
+
+import sys
+sys.path.append("/content/Pig-weight-calculation-by-Mask-R-CNN-Keras-and-TensorFlow/mrcnn")
+from m_rcnn import *
+
+"""## **3. Image Dataset**
+
+Load your images and annotated dataset
+
+"""
+
+images_path = '/content/Pig-weight-calculation-by-Mask-R-CNN-Keras-and-TensorFlow/images'
+annotations_path = "/content/Pig-weight-calculation-by-Mask-R-CNN-Keras-and-TensorFlow/annotations2.json"
+dataset_train = load_image_dataset(os.path.join(annotations_path), images_path, "train")
+dataset_val = load_image_dataset(os.path.join(annotations_path), images_path, "val")
+class_number = dataset_train.count_classes()
+print('Train: %d' % len(dataset_train.image_ids))
+print('Validation: %d' % len(dataset_val.image_ids))
+print("Classes: {}".format(class_number))
+
+"""Load image samples"""
+
+display_image_samples(dataset_train)
+
+"""## **4. Training**
+
+Train Mask RCNN on your custom Dataset.
+"""
+
+# Load Configuration
+config = CustomConfig(class_number)
+#config.display()
+model = load_training_model(config)
+
+# Start Training
+# This operation might take a long time.
+train_head(model, dataset_train, dataset_train, config)
+
+"""## **5. Save your model**
+
+Save your model training in "mask_rcnn_shapes.h5"
+"""
+
+model_path = os.path.join(ROOT_DIR, "mask_rcnn_shapes.h5")
+model.keras_model.save_weights(model_path)
+#Export results
+from google.colab import files
+files.download("/content/Pig-weight-calculation-by-Mask-R-CNN-Keras-and-TensorFlow/mask_rcnn_shapes.h5")
+
+"""## **6. Detection (test your model on a random image)**"""
+
+# Load Test Model
+# The latest trained model will be loaded
+test_model, inference_config = load_test_model(class_number)
+
+# Test on a random image
+test_random_image(test_model, dataset_val, inference_config)
+
+"""## **7. Run Mask-RCNN on Images**
+
+You can load here the image and extract the mask using Mask-RCNN
+
+"""
+
+import os
+import glob
+import pandas as pd
+import cv2
+from google.colab.patches import cv2_imshow
+from visualize import *
+
+files_images=[]
+path_images = '/content/Pig-weight-calculation-by-Mask-R-CNN-Keras-and-TensorFlow/images'
+images = pd.DataFrame([f for f in glob.glob(path_images + "**/*.jpg", recursive=True)], columns = ['Name'])
+n_rows = images.shape[0]
+
+"""Here, you can calculate the area. Only assign a value if you know the actual size of objects."""
+
+RATIO_PIXEL_TO_CM_h = 2.4 # 2.4 pixels are 1cm
+RATIO_PIXEL_TO_CM_w = 2.9
+RATIO_PIXEL_TO_SQUARE_CM = RATIO_PIXEL_TO_CM_h * RATIO_PIXEL_TO_CM_w
+results = {}
+
+def detect_contours_maskrcnn(model, img):
+    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    results = model.detect([img_rgb])
+    r = results[0]
+    object_count = len(r["class_ids"])
+    
+    objects_ids = []
+    objects_contours = []
+    bboxes = []
+    for i in range(object_count):
+        # 1. Class ID
+        class_id = r["class_ids"][i]
+        # 2. Boxes
+        box = r["rois"][i]
+        
+        # 3. Mask
+        mask = r["masks"][:, :, i]
+        contours = get_mask_contours(mask)
+        bboxes.append(box)
+        objects_contours.append(contours[0])
+        objects_ids.append(class_id)
+    return objects_ids, bboxes, objects_contours
+
+"""Now, let's create a loop for each image and calculate the size of the pigs."""
+
+for i in range(n_rows):
+  # Save image in img
+  img = cv2.imread(images.loc[i, "Name"])
+  #cv2_imshow(img)
+  # Created box
+  img_box = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
+  #Colors to class
+  class_names = ["BG", "green", "blue", "light blue", "pink"]
+  colors = random_colors(len(class_names))
+
+  # Get objects mask
+  class_ids, boxes, masks = detect_contours_maskrcnn(test_model, img)
+
+  for class_id, box, object_contours in zip(class_ids, boxes, masks):
+      # 1. Creted polylines Box to calculate size of the pigs
+      y1, x1, y2, x2 = box
+      #cv2.rectangle(img, (x1, y1), (x2, y2), colors[class_id], 15)
+      cv2.polylines(img, [object_contours], True, colors[class_id], 2)
+      img = draw_mask(img, [object_contours], colors[class_id])
+
+      # 2. Calculate area
+      area_px = cv2.contourArea(object_contours)
+      area_cm = round(area_px / RATIO_PIXEL_TO_SQUARE_CM, 2)
+
+      # 3. Calculate perimeter
+      perimeter_px = cv2.arcLength(object_contours, True)
+      perimeter_cm = round((perimeter_px *  600)/1640,2)
+
+      # 4. Calculate length
+      rect = cv2.minAreaRect(object_contours)
+      box = cv2.boxPoints(rect)
+      box = np.int0(box)
+      cv2.drawContours(img, [box], 0, (0,255,0), 2) # this was mostly for debugging you may omit
+      (x, y), (w, h), angle = rect
+      
+      # 5. Get Width and Height of the Objects by applying the Ratio pixel to cm
+
+      object_width = round(w* 600/ 1640,2)
+      object_height = round(h *600/ 1640 ,2)
+
+      # 6. Add informations in the images
+      cv2.putText(img, "Nome: {}".format(images.loc[i, "Name"][74:-4]), (0, 300), cv2.FONT_HERSHEY_PLAIN, 2, colors[class_id], 2)
+      cv2.putText(img, "A: {}cm^2".format(round(area_px,2)), (0, 250), cv2.FONT_HERSHEY_PLAIN, 2, colors[class_id], 2)
+      cv2.putText(img, "P: {}cm".format(round(perimeter_px,2)), (0, 200), cv2.FONT_HERSHEY_PLAIN,2, colors[class_id], 2)
+      cv2.putText(img, "C: {}cm".format(round(h,2)), (0, 150), cv2.FONT_HERSHEY_PLAIN, 2, colors[class_id], 2)
+      cv2.putText(img, "L: {}cm".format(round(w,2)), (0, 100), cv2.FONT_HERSHEY_PLAIN, 2, colors[class_id], 2)
+      
+      # 7. Save informations in the results
+      Nome = images.loc[i, "Name"][74:-4]
+      results[i] = {"Nome": Nome,
+                    "Area_cm": area_cm,
+                    "Perimetro_cm":perimeter_cm,
+                    "Largura_cm": object_height,
+                    "Comprimento_cm":object_width,
+                    "Area_px": area_px,
+                    "Perimetro_px":perimeter_px,
+                    "Largura_px": w,
+                    "Comprimento_px":h}
+  # Plot image
+  cv2_imshow(img)
+
+"""Save results """
+
+results =  pd.DataFrame(data=results)
+results
+writer = pd.ExcelWriter('results.xlsx')
+results.to_excel(writer,'Sheet1')
+writer.save()
+
+#Export results
+from google.colab import files
+files.download("/content/results.xlsx")
+
+"""### **8. Inference Mask-RCNN on Images**
+
+You can load here the image and extract the mask using Mask-RCNN
+"""
+
+import os
+import pandas as pd
+resultsfiles_images=[]
+path_images = '/content/Pig-weight-calculation-by-Mask-R-CNN-Keras-and-TensorFlow/images'
+images = pd.DataFrame(os.listdir(path_images), columns = ['Name'])
+n_rows = images.shape[0]
+
+import cv2
+from google.colab.patches import cv2_imshow
+
+RATIO_PIXEL_TO_CM_h = 2.4 # 2.4 pixels are 1cm
+RATIO_PIXEL_TO_CM_w = 2.9
+RATIO_PIXEL_TO_SQUARE_CM = RATIO_PIXEL_TO_CM_h * RATIO_PIXEL_TO_CM_w
+
+results = {}
+
+"""We can load the calculated weights and now if we have more images we can insert them into the inference model to get their measurements."""
+
+# Model inference
+test_model, inference_config = load_inference_model(1, "/content/Pig-weight-calculation-by-Mask-R-CNN-Keras-and-TensorFlow/mask_rcnn_shapes.h5")
+
+for i in range(n_rows):
+
+  img = cv2.imread(os.path.join(path_images, images.loc[i, "Name"]))
+
+  img_box = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
+  class_names = ["BG", "green", "blue", "light blue", "pink"]
+  colors = random_colors(len(class_names))
+
+  image = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+
+  # Detect results
+  r = test_model.detect([image])[0]
+  colors = random_colors(80)
+
+  # Get Coordinates and show it on the image
+  object_count = len(r["class_ids"])
+  for j in range(object_count):
+      # 1. Mask
+      mask = r["masks"][:, :, j]
+      contours = get_mask_contours(mask)
+      for cnt in contours:
+          cv2.polylines(img, [cnt], True, colors[j], 2)
+          img = draw_mask(img, [cnt], colors[j])
+          # 2. Calculate area
+          area_px = cv2.contourArea(cnt)
+          area_cm = round(area_px / RATIO_PIXEL_TO_SQUARE_CM, 2)
+
+          # 3. Calculate perimeter
+          perimeter_px = cv2.arcLength(cnt, True)
+          perimeter_cm = round((perimeter_px *  600)/1640,2)
+
+          # 4. Calculate length
+          rect = cv2.minAreaRect(cnt)
+          box = cv2.boxPoints(rect)
+          box = np.int0(box)
+          cv2.drawContours(img, [box], 0, colors[j], 2) # this was mostly for debugging you may omit
+          (x, y), (w, h), angle = rect
+          if (w > h):
+              C_px = object_width
+              L_px = object_height
+          if (h > w):
+              L_px = object_width
+              C_px = object_height
+          
+          object_width = round(w* 600/ 1640,2)
+          object_height = round(h *600/ 1640 ,2)
+
+          if (object_width > object_height):
+              C_cm = object_width
+              L_cm = object_height
+          if (object_height > object_width):
+              L_cm = object_width
+              C_cm = object_height
+          
+          # 5. write in image
+          peso = (-1.49560824863731 + 0.00284305110419364*area_px)
+          cv2.putText(img, "Nome: {}".format(images.loc[i, "Name"]), (0, 350), cv2.FONT_HERSHEY_PLAIN, 2, colors[j], 2)
+          cv2.putText(img, "Peso predito: {}".format(round(peso,2)), (0, 300), cv2.FONT_HERSHEY_PLAIN, 2, colors[j], 2)
+          cv2.putText(img, "A: {}cm^2".format(round(area_cm,2)), (0, 250), cv2.FONT_HERSHEY_PLAIN, 2, colors[j], 2)
+          cv2.putText(img, "P: {}cm".format(round(perimeter_cm,2)), (0, 200), cv2.FONT_HERSHEY_PLAIN,2, colors[j], 2)
+          cv2.putText(img, "C: {}cm".format(round(C_cm,2)), (0, 150), cv2.FONT_HERSHEY_PLAIN, 2, colors[j], 2)
+          cv2.putText(img, "L: {}cm".format(round(L_cm,2)), (0, 100), cv2.FONT_HERSHEY_PLAIN, 2, colors[j], 2)
+
+          results[i] = {"Nome": images.loc[i, "Name"],
+                        "Area_cm": area_cm,
+                        "Perimetro_cm":perimeter_cm,
+                        "Largura_cm": L_cm,
+                        "Comprimento_cm":C_cm,
+                        "Area_px": area_px,
+                        "Perimetro_px":perimeter_px,
+                        "Largura_px": L_px,
+                        "Comprimento_px":C_px,
+                        "Peso": peso}
+
+          cv2_imshow(img)
+
+results =  pd.DataFrame(data=results)
+results
+writer = pd.ExcelWriter('results_inference.xlsx')
+results.to_excel(writer,'Sheet1')
+writer.save()
+
+#Export results
+from google.colab import files
+files.download("/content/results_inference.xlsx")