app.py

import gradio as gr
import cv2
import requests
import os
import numpy as np
from math import atan2, cos, sin, sqrt, pi
from ultralytics import YOLO

def drawAxis(img, p_, q_, color, scale):
  p = list(p_)
  q = list(q_)
 
  ## [visualization1]
  angle = atan2(p[1] - q[1], p[0] - q[0]) # angle in radians
  hypotenuse = sqrt((p[1] - q[1]) * (p[1] - q[1]) + (p[0] - q[0]) * (p[0] - q[0]))
 
  # Here we lengthen the arrow by a factor of scale
  q[0] = p[0] - scale * hypotenuse/2 * cos(angle)
  q[1] = p[1] - scale * hypotenuse/2 * sin(angle)
  cv2.line(img, (int(p[0]), int(p[1])), (int(q[0]), int(q[1])), color, 3, cv2.LINE_AA)
 
  # create the arrow hooks
  p[0] = q[0] + 9 * cos(angle + pi / 4)
  p[1] = q[1] + 9 * sin(angle + pi / 4)
  cv2.line(img, (int(p[0]), int(p[1])), (int(q[0]), int(q[1])), color, 3, cv2.LINE_AA)
 
  p[0] = q[0] + 9 * cos(angle - pi / 4)
  p[1] = q[1] + 9 * sin(angle - pi / 4)
  cv2.line(img, (int(p[0]), int(p[1])), (int(q[0]), int(q[1])), color, 3, cv2.LINE_AA)
  ## [visualization1]
 

def getOrientation(pts, img):
  ## [pca]
  # Construct a buffer used by the pca analysis
  sz = len(pts)
  data_pts = np.empty((sz, 2), dtype=np.float64)
  for i in range(data_pts.shape[0]):
    data_pts[i,0] = pts[i,0,0]
    data_pts[i,1] = pts[i,0,1]
 
  # Perform PCA analysis
  mean = np.empty((0))
  mean, eigenvectors, eigenvalues = cv2.PCACompute2(data_pts, mean)
 
  # Store the center of the object
  cntr = (int(mean[0,0]), int(mean[0,1]))
  ## [pca]
 
  ## [visualization]
  # Draw the principal components
  cv2.circle(img, cntr, 3, (255, 0, 255), 10)
  p1 = (cntr[0] + 0.02 * eigenvectors[0,0] * eigenvalues[0,0], cntr[1] + 0.02 * eigenvectors[0,1] * eigenvalues[0,0])
  p2 = (cntr[0] - 0.02 * eigenvectors[1,0] * eigenvalues[1,0], cntr[1] - 0.02 * eigenvectors[1,1] * eigenvalues[1,0])
  drawAxis(img, cntr, p1, (255, 255, 0), 1)
  drawAxis(img, cntr, p2, (255, 255, 0), 4)
 
  angle = atan2(eigenvectors[0,1], eigenvectors[0,0]) # orientation in radians
  ## [visualization]
  angle_deg = -(int(np.rad2deg(angle))-180) % 180
 
#   # Label with the rotation angle
#   label = str(int(np.rad2deg(angle))) + " deg"
#   textbox = cv2.rectangle(img, (cntr[0]+60, cntr[1]-25), (cntr[0] + 150, cntr[1] + 10), (255,255,255), -1)
#   cv2.putText(img, label, (cntr[0]+60, cntr[1]), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0,0,0), 1, cv2.LINE_AA)

  return angle_deg

file_urls = [
    'https://github.com/lucarei/orientation-detection-robotic-grasping/assets/22428774/cefd9731-c57c-428b-b401-fd54a8bd0a95',
    'https://github.com/lucarei/orientation-detection-robotic-grasping/assets/22428774/acbad76a-33f9-4028-b012-4ece5998c272',
    'https://github.com/lucarei/orientation-detection-robotic-grasping/assets/22428774/ce8a0fb9-99ea-4952-bcc4-3afa023066d9',
    'https://dl.dropboxusercontent.com/scl/fi/flbf7vvoxgzoe9adovadm/video-540p.mp4?dl=0&rlkey=jbecmpu727q7yirvquma9m7w2'
]

def download_file(url, save_name):
    url = url
    if not os.path.exists(save_name):
        file = requests.get(url)
        open(save_name, 'wb').write(file.content)

for i, url in enumerate(file_urls):
    if 'mp4' in file_urls[i]:
        download_file(
            file_urls[i],
            f"video.mp4"
        )
    else:
        download_file(
            file_urls[i],
            f"image_{i}.jpg"
        )

model = YOLO('best.pt')
path  = [['image_0.jpg'], ['image_1.jpg'], ['image_2.jpg']]
video_path = [['video.mp4']]

def show_preds_image(image_path):
    image = cv2.imread(image_path)
    #resize image (optional)
    img_res_toshow = cv2.resize(image, None, fx= 0.5, fy= 0.5, interpolation= cv2.INTER_LINEAR)
    height=img_res_toshow.shape[0]
    width=img_res_toshow.shape[1]
    dim=(width,height)

    outputs = model.predict(source=img_res_toshow)
    boxes = outputs[0].boxes.cls
    class_list = []
    for class_n in boxes.cpu().numpy():
        class_list.append(outputs[0].names[class_n])
    angle_list = []
    for object in range(len(outputs[0].masks.masks)):
        #obtain BW image
        bw=(outputs[0].masks.masks[object].cpu().numpy() * 255).astype("uint8")
        #BW image with same dimention of initial image
        bw=cv2.resize(bw, dim, interpolation = cv2.INTER_AREA)
        img=img_res_toshow
        contours, _ = cv2.findContours(bw, cv2.RETR_LIST, cv2.CHAIN_APPROX_NONE)
        for i, c in enumerate(contours):
            # Calculate the area of each contour
            area = cv2.contourArea(c)
            
            # Ignore contours that are too small or too large
            if area < 2500 or 500000 < area:
                continue
            
            # Draw each contour only for visualisation purposes
            cv2.drawContours(img, contours, i, (0, 0, 255), 2)
            
            # Find the orientation of each shape
            angle_deg = getOrientation(c, img)
            angle_list.append(angle_deg)

    # Create a text string for the orientation of the objects and the class of the objects
    text = ""
    for i in range(len(angle_list)):
        text = text + "Object " + str(i+1) + ": " + str(angle_list[i]) + " deg, " + class_list[i] + "\n"
        
    results = outputs[0].cpu().numpy()
    for i, det in enumerate(results.boxes.xyxy):
        cv2.rectangle(
            img,
            (int(det[0]), int(det[1])),
            (int(det[2]), int(det[3])),
            color=(255, 0, 0),
            thickness=2,
            lineType=cv2.LINE_AA
        )
    return cv2.cvtColor(img, cv2.COLOR_BGR2RGB), text

inputs_image = [
    gr.components.Image(type="filepath", label="Input Image"),
]
outputs_image = [
    gr.components.Image(type="numpy", label="Output Image"),
    gr.outputs.Textbox(label="Orientation Angle")
]
interface_image = gr.Interface(
    fn=show_preds_image,
    inputs=inputs_image,
    outputs=outputs_image,
    title="Trash Detection with Orientation",
    examples=path,
    cache_examples=False,
)

def show_preds_video(video_path):
    cap = cv2.VideoCapture(video_path)
    while(cap.isOpened()):
        ret, frame = cap.read()
        if ret:
            frame_copy = frame.copy()
            #resize image (optional)
            img_res_toshow = cv2.resize(frame_copy, None, fx= 0.5, fy= 0.5, interpolation= cv2.INTER_LINEAR)
            height=img_res_toshow.shape[0]
            width=img_res_toshow.shape[1]
            dim=(width,height)
            outputs = model.predict(source=img_res_toshow)
            for object in range(len(outputs[0].masks.masks)):
                #obtain BW image
                bw=(outputs[0].masks.masks[object].cpu().numpy() * 255).astype("uint8")
                #BW image with same dimention of initial image
                bw=cv2.resize(bw, dim, interpolation = cv2.INTER_AREA)
                img=img_res_toshow
                contours, _ = cv2.findContours(bw, cv2.RETR_LIST, cv2.CHAIN_APPROX_NONE)
                for i, c in enumerate(contours):
                    # Calculate the area of each contour
                    area = cv2.contourArea(c)
                    
                    # Ignore contours that are too small or too large
                    if area < 2500 or 500000 < area:
                        continue
                    
                    # Draw each contour only for visualisation purposes
                    cv2.drawContours(img, contours, i, (0, 0, 255), 2)
                    
                    # Find the orientation of each shape
                    angle_deg = getOrientation(c, img)

            results = outputs[0].cpu().numpy()
            for i, det in enumerate(results.boxes.xyxy):
                cv2.rectangle(
                    img,
                    (int(det[0]), int(det[1])),
                    (int(det[2]), int(det[3])),
                    color=(255, 0, 0),
                    thickness=2,
                    lineType=cv2.LINE_AA
                )
            yield cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

inputs_video = [
    gr.components.Video(type="filepath", label="Input Video"),

]
outputs_video = [
    gr.components.Image(type="numpy", label="Output Image"),
]
interface_video = gr.Interface(
    fn=show_preds_video,
    inputs=inputs_video,
    outputs=outputs_video,
    title="Trash Detection with Orientation",
    examples=video_path,
    cache_examples=False,
)

gr.TabbedInterface(
    [interface_image, interface_video],
    tab_names=['Image inference', 'Video inference']
).queue().launch()