""" Use right click for initial global point selection. Can always be overwritten by selecting new points. Use left click to add tracking points. New left clicks will add more points. To remove previous ones, use middle click. Pressing next button will automatically save current progress and will go to the next video in the same directory. """ import os import sys import cv2 import json import numpy as np import pandas as pd from natsort import natsorted from dataclasses import dataclass, field from typing import List, Tuple, Dict, Optional from PyQt5 import QtCore, QtWidgets, QtGui # INPUT_FRAME_DIR = "/path/to/Cataract-1K/video_frames_jpg" # OUTPUT_ANN_DIR = "/path/to/ann/Cataract-1K/ann_glob_pretracking_points" INPUT_FRAME_DIR = "/path/to/Cataracts-50/video_frames_jpg" OUTPUT_ANN_DIR = "/path/to/ann/Cataracts-50/ann_glob_pretracking_points" # Switch this to None if normal annotation POST_REFINEMENT_ANN = None # POST_REFINEMENT_ANN = "/path/to/ann/Cataracts-50/ann_glob_posttracking_refinement.csv" @dataclass class TrackingPoints: points: List[Tuple[int, int]] = field(default_factory=list) def undo(self): self.points = [p for p in self.points[:-1]] def clear(self): self.points = [] @dataclass class PointObject: name: str frame_idx: int = 0 zoom_level: str = None global_point: Tuple[int, int] = None tracking_points: TrackingPoints = field(default_factory=TrackingPoints) class Point_Selector(QtWidgets.QMainWindow): def __init__(self): super().__init__() self.setWindowTitle("Select Global and Tracking Points") self.setGeometry(50, 50, 650, 385) self.input_frame_dir = INPUT_FRAME_DIR self.output_ann_path = OUTPUT_ANN_DIR self.post_refinement_ann_path = POST_REFINEMENT_ANN self.frame_size = (512, 288) self.video_path = None self.init_ui() def init_obj(self): video_name = os.path.basename(self.video_path) self.obj = PointObject(name=video_name) if self.btn_zoom_level.checkedButton() is not None: zoom_text = self.btn_zoom_level.checkedButton().text() self.select_zoom_level(zoom_text) def init_ui(self): central = QtWidgets.QWidget() self.setCentralWidget(central) layout = QtWidgets.QHBoxLayout(central) display_layout = QtWidgets.QVBoxLayout() button_layout = QtWidgets.QVBoxLayout() self.video_label = QtWidgets.QLabel() self.video_label.setFixedSize(self.frame_size[0], self.frame_size[1]) self.video_label.setScaledContents(True) self.video_label.installEventFilter(self) self.current_qimage = None title_layout = QtWidgets.QHBoxLayout() self.vid_title_status = QtWidgets.QLabel("") title_layout.addWidget(self.vid_title_status) title_layout.addSpacing(25) self.text_box = QtWidgets.QLineEdit() self.text_box.setPlaceholderText("Enter Frame Index...") self.text_box.returnPressed.connect(self.on_textbox_entered) title_layout.addWidget(self.text_box) display_layout.addLayout(title_layout) display_layout.addWidget(self.video_label, alignment=QtCore.Qt.AlignCenter) self.frame_status = QtWidgets.QLabel("") display_layout.addWidget(self.frame_status) self.status = QtWidgets.QLabel("") display_layout.addWidget(self.status) button_layout.addStretch() button_layout.addSpacing(-25) btn_load = QtWidgets.QPushButton("Load Video") btn_load.clicked.connect(self.select_video) button_layout.addWidget(btn_load) self.btn_load_next = QtWidgets.QPushButton("Load Next") self.btn_load_next.clicked.connect(self.next_video) self.btn_load_next.setEnabled(False) button_layout.addWidget(self.btn_load_next) self.btn_load_prev = QtWidgets.QPushButton("Load Previous") self.btn_load_prev.clicked.connect(self.previous_video) self.btn_load_prev.setEnabled(False) button_layout.addWidget(self.btn_load_prev) button_layout.addSpacing(20) self.btn_clear_all = QtWidgets.QPushButton("Clear All") self.btn_clear_all.clicked.connect(self.clear_annotations) self.btn_clear_all.setEnabled(False) button_layout.addWidget(self.btn_clear_all) button_layout.addSpacing(20) self.btn_zoom_level = QtWidgets.QButtonGroup(self) self.btn_zoom_level.setExclusive(True) btn_zoom_low = QtWidgets.QPushButton("Zoom: Low") btn_zoom_low.clicked.connect(lambda checked: self.select_zoom_level(btn_zoom_low.text())) btn_zoom_low.setCheckable(True) self.btn_zoom_level.addButton(btn_zoom_low) button_layout.addWidget(btn_zoom_low) btn_zoom_high = QtWidgets.QPushButton("Zoom: High") btn_zoom_high.clicked.connect(lambda checked: self.select_zoom_level(btn_zoom_high.text())) btn_zoom_high.setCheckable(True) self.btn_zoom_level.addButton(btn_zoom_high) button_layout.addWidget(btn_zoom_high) button_layout.addSpacing(20) self.btn_next_frame = QtWidgets.QPushButton("Frame: Next") self.btn_next_frame.clicked.connect(lambda: self.change_frame(+4)) self.btn_next_frame.setEnabled(False) button_layout.addWidget(self.btn_next_frame) self.btn_prev_frame = QtWidgets.QPushButton("Frame: Previous") self.btn_prev_frame.clicked.connect(lambda: self.change_frame(-4)) self.btn_prev_frame.setEnabled(False) button_layout.addWidget(self.btn_prev_frame) self.btn_first_frame = QtWidgets.QPushButton("Frame: First") self.btn_first_frame.clicked.connect(lambda: self.change_frame(-100000)) self.btn_first_frame.setEnabled(False) button_layout.addWidget(self.btn_first_frame) button_layout.addStretch() layout.addLayout(display_layout) layout.addLayout(button_layout) def select_video(self): if self.post_refinement_ann_path is None: path = QtWidgets.QFileDialog.getExistingDirectory(self, "Select Folder", "") if not path: QtWidgets.QMessageBox.critical(self, "Error", "No folder selected") return self.video_path = path self.init_obj() self.goto_frame() else: self.post_refinement_df = pd.read_csv(self.post_refinement_ann_path, skipinitialspace=True) self.video_path = os.path.join(self.input_frame_dir, self.post_refinement_df.iloc[0]['video_id']) self.status.setText("Total Frames:" + str(len(os.listdir(self.video_path)))) self.init_obj() self.obj.frame_idx = int(self.post_refinement_df.iloc[0]['frame_idx']) self.goto_frame() self.btn_next_frame.setEnabled(True) self.btn_prev_frame.setEnabled(True) self.btn_first_frame.setEnabled(True) self.btn_clear_all.setEnabled(True) self.btn_load_next.setEnabled(True) self.btn_load_prev.setEnabled(True) def find_video(self, go_previous=False): parent_path = os.path.dirname(self.video_path) video_name = os.path.basename(self.video_path) if self.post_refinement_ann_path is None: video_list = natsorted(os.listdir(parent_path)) else: video_list = self.post_refinement_df['video_id'].unique().tolist() video_idx = video_list.index(video_name) video_idx = video_idx - 1 if go_previous else video_idx + 1 if video_idx >= len(video_list) or video_idx < 0: QtWidgets.QMessageBox.information(self, "Info", "No more videos in the directory") return if self.post_refinement_ann_path is None: self.video_path = os.path.join(parent_path, video_list[video_idx]) self.init_obj() else: self.video_path = os.path.join(self.input_frame_dir, self.post_refinement_df.iloc[video_idx]['video_id']) self.status.setText("Total Frames:" + str(len(os.listdir(self.video_path)))) self.init_obj() self.obj.frame_idx = int(self.post_refinement_df.iloc[video_idx]['frame_idx']) self.frame_status.setText(f"Total Tracking Points: {len(self.obj.tracking_points.points)} / 40") self.goto_frame() def next_video(self): if not self.export_annotations(): return self.find_video(go_previous=False) def previous_video(self): self.find_video(go_previous=True) def on_textbox_entered(self): text = self.text_box.text() if text.isdigit(): frame_idx = int(text) self.obj.frame_idx = frame_idx self.goto_frame() else: self.status.setText(f"Please enter a valid frame index (integer).") def change_frame(self, delta: int): new_idx = self.obj.frame_idx + delta if new_idx >= 0: self.obj.frame_idx = new_idx else: self.obj.frame_idx = 0 self.goto_frame() def goto_frame(self): frame_idx = str(self.obj.frame_idx).zfill(10) frame_path = os.path.join(self.video_path, frame_idx+".jpg") frame_rgb = cv2.imread(frame_path) frame_rgb = cv2.cvtColor(frame_rgb, cv2.COLOR_BGR2RGB) frame_rgb = cv2.resize(frame_rgb, self.frame_size, interpolation=cv2.INTER_AREA) frame_rgb = self.render_overlays(frame_rgb) self.video_label_set_frame(frame_rgb) self.vid_title_status.setText(f"Selected Video: {os.path.basename(self.video_path)} | Frame Index: {frame_idx}") def clear_annotations(self): self.init_obj() self.goto_frame() self.frame_status.setText(f"Total Tracking Points: {len(self.obj.tracking_points.points)} / 40") self.status.setText(f"Cleared All Annotations") def export_annotations(self): if self.obj.global_point is not None and len(self.obj.tracking_points.points) > 40 and self.obj.zoom_level is not None: ann_data = { "name": self.obj.name, "zoom_level": self.obj.zoom_level, "frames": [{ "frame_idx": self.obj.frame_idx, "global_point": self.obj.global_point, "tracking_points": self.obj.tracking_points.points }]} ann_file_path = os.path.join(self.output_ann_path, f"{self.obj.name}.json") # Read first the json, if exists only add for new idx or replace existing idx if os.path.exists(ann_file_path): with open(ann_file_path, "r") as f: old_ann_data = json.load(f) for frame in old_ann_data["frames"]: if frame["frame_idx"] != self.obj.frame_idx: ann_data["frames"].append(frame) # Sort frames by frame idx ann_data["frames"] = sorted(ann_data["frames"], key=lambda x: x["frame_idx"]) with open(ann_file_path, 'w') as f: json.dump(ann_data, f, indent=4) self.status.setText(f"Saving Annotations Before the Loading Next Video") return True else: self.status.setText(f"No Sufficient Annotations to Save") return False def select_zoom_level(self, zoom_level: str): if hasattr(self, "obj"): if zoom_level == "Zoom: Low": self.obj.zoom_level = "low" elif zoom_level == "Zoom: High": self.obj.zoom_level = "high" return def render_overlays(self, frame_rgb: np.ndarray) -> np.ndarray: canvas = frame_rgb.copy() if canvas.dtype != np.uint8: canvas = (canvas * 255).astype(np.uint8) if canvas.max() <= 1.0 else canvas.astype(np.uint8) h, w, _ = canvas.shape canvas_f = canvas.astype(np.float32) if self.obj.global_point is not None: x, y = self.obj.global_point cv2.circle(canvas_f, (int(x), int(y)), 3, (255,0,0), -1, lineType=cv2.LINE_AA) cv2.circle(canvas_f, (int(x), int(y)), 100, (255,0,0), 2, lineType=cv2.LINE_AA) for p in self.obj.tracking_points.points: x, y = p cv2.circle(canvas_f, (int(x), int(y)), 3, (20,150,255), -1, lineType=cv2.LINE_AA) canvas_out = np.clip(canvas_f, 0, 255).astype(np.uint8) return canvas_out def video_label_set_frame(self, rgb_frame: np.ndarray): if rgb_frame is None: return h, w, ch = rgb_frame.shape bytes_per_line = ch * w qimg = QtGui.QImage(rgb_frame.data, w, h, bytes_per_line, QtGui.QImage.Format_RGB888) self.current_qimage = qimg.copy() pix = QtGui.QPixmap.fromImage(self.current_qimage) self.video_label.setPixmap(pix.scaled(self.video_label.size(), QtCore.Qt.KeepAspectRatio, QtCore.Qt.SmoothTransformation)) def eventFilter(self, obj, event): if obj == self.video_label: if event.type() == QtCore.QEvent.MouseButtonPress: pos = event.pos() x, y = pos.x(), pos.y() if event.button() == QtCore.Qt.RightButton: self.obj.global_point = (x, y) self.status.setText(f"Global Point is Set") elif event.button() == QtCore.Qt.LeftButton: self.obj.tracking_points.points.append((x, y)) self.status.setText(f"Added New Tracking Point") elif event.button() == QtCore.Qt.MiddleButton: self.obj.tracking_points.undo() self.status.setText(f"Removed Previous Tracking Point") self.frame_status.setText(f"Total Tracking Points: {len(self.obj.tracking_points.points)} / 40") self.goto_frame() return super().eventFilter(obj, event) if __name__ == '__main__': app = QtWidgets.QApplication(sys.argv) win = Point_Selector() win.show() sys.exit(app.exec_())