import os import re import shutil import threading import time import tkinter as tk from tkinter import filedialog, messagebox, simpledialog from tkinter.scrolledtext import ScrolledText import cv2 from PIL import Image, ImageTk class UniversalAnnotationTool: def __init__(self, root): self.root = root self.root.title("通用标注与裁剪工具") self.root.geometry("1600x920") self.image = None self.photo = None self.file_path = None self.folder_path = None self.image_list = [] self.current_image_index = 0 self.coordinates = {} self.current_shape = None self.mode = "polygon" self.points = [] self.lines = [] self.point_labels = [] self.box_preview = None self.box_preview_text = None self.box_start = None self.box_coords = None self.box_shape_id = None self.box_text_id = None self.restored_points = [] self.restored_lines = [] self.restored_text_labels = [] self.restored_rectangles = [] self.rtsp_url = "" self.cap = None self.rtsp_thread = None self._build_layout() self._bind_events() self.enable_polygon_mode() def _build_layout(self): main_frame = tk.Frame(self.root) main_frame.pack(fill="both", expand=True) left_frame = tk.Frame(main_frame) left_frame.pack(side="left", fill="both", expand=True, padx=10, pady=10) canvas_frame = tk.Frame(left_frame) canvas_frame.pack(fill="both", expand=True) v_scrollbar = tk.Scrollbar(canvas_frame, orient="vertical") v_scrollbar.pack(side="right", fill="y") h_scrollbar = tk.Scrollbar(canvas_frame, orient="horizontal") h_scrollbar.pack(side="bottom", fill="x") self.canvas = tk.Canvas( canvas_frame, bg="white", width=1100, height=760, yscrollcommand=v_scrollbar.set, xscrollcommand=h_scrollbar.set, ) self.canvas.pack(side="left", fill="both", expand=True) v_scrollbar.config(command=self.canvas.yview) h_scrollbar.config(command=self.canvas.xview) btn_row1 = tk.Frame(left_frame) btn_row1.pack(fill="x", pady=(10, 4)) btn_row2 = tk.Frame(left_frame) btn_row2.pack(fill="x", pady=4) btn_row3 = tk.Frame(left_frame) btn_row3.pack(fill="x", pady=4) tk.Button(btn_row1, text="上传文件", command=self.upload_file, width=12).pack(side="left", padx=3) tk.Button(btn_row1, text="选择文件夹", command=self.select_folder, width=12).pack(side="left", padx=3) tk.Button(btn_row1, text="上一张", command=self.prev_image, width=12).pack(side="left", padx=3) tk.Button(btn_row1, text="下一张", command=self.next_image, width=12).pack(side="left", padx=3) tk.Button(btn_row1, text="保存坐标", command=self.save_coordinates, width=12).pack(side="left", padx=3) tk.Button(btn_row1, text="重置", command=self.reset, width=12).pack(side="left", padx=3) tk.Button(btn_row2, text="多边形模式", command=self.enable_polygon_mode, width=12).pack(side="left", padx=3) tk.Button(btn_row2, text="框选模式", command=self.enable_box_mode, width=12).pack(side="left", padx=3) tk.Button(btn_row2, text="撤回", command=self.undo_last_point, width=12).pack(side="left", padx=3) tk.Button(btn_row2, text="恢复输入", command=self.restore_coordinates, width=12).pack(side="left", padx=3) tk.Button(btn_row2, text="批量回显", command=self.batch_restore_coordinates, width=12).pack(side="left", padx=3) tk.Button(btn_row2, text="裁剪", command=self.crop_image, width=12).pack(side="left", padx=3) tk.Button(btn_row2, text="压缩当前", command=self.compress_image, width=12).pack(side="left", padx=3) tk.Button(btn_row2, text="批量压缩", command=self.compress_all_images, width=12).pack(side="left", padx=3) tk.Button(btn_row3, text="连接 RTSP", command=self.connect_rtsp, width=12).pack(side="left", padx=3) tk.Button(btn_row3, text="截图 RTSP", command=self.capture_rtsp_frame, width=12).pack(side="left", padx=3) tk.Button(btn_row3, text="断开 RTSP", command=self.disconnect_rtsp, width=12).pack(side="left", padx=3) tk.Button(btn_row3, text="视频抽帧", command=self.extract_video_frames, width=12).pack(side="left", padx=3) tk.Button(btn_row3, text="整理标注图", command=self.collect_labeled_images, width=12).pack(side="left", padx=3) self.status_label = tk.Label(left_frame, text="状态: 就绪", anchor="w", fg="darkgreen") self.status_label.pack(fill="x", pady=(6, 0)) right_frame = tk.Frame(main_frame) right_frame.pack(side="right", fill="y", padx=10, pady=10) tk.Label(right_frame, text="图片列表", font=("Arial", 12, "bold")).pack(anchor="w") self.image_listbox = tk.Listbox(right_frame, width=34, height=24) self.image_listbox.pack(fill="y", pady=(4, 8)) self.image_listbox.bind("<>", self.on_image_select) search_frame = tk.Frame(right_frame) search_frame.pack(fill="x", pady=(0, 8)) tk.Label(search_frame, text="搜索").pack(side="left") self.search_entry = tk.Entry(search_frame) self.search_entry.pack(side="left", fill="x", expand=True, padx=(6, 0)) self.search_entry.bind("", self.search_image) tk.Label(right_frame, text="坐标 / 标注输入").pack(anchor="w") self.coords_text = ScrolledText(right_frame, width=38, height=10) self.coords_text.pack(fill="both", pady=(4, 8)) crop_frame = tk.Frame(right_frame) crop_frame.pack(fill="x", pady=(0, 8)) tk.Label(crop_frame, text="裁剪坐标").pack(side="left") self.crop_entry = tk.Entry(crop_frame) self.crop_entry.pack(side="left", fill="x", expand=True, padx=(6, 0)) tk.Label(right_frame, text="说明").pack(anchor="w") self.help_text = tk.Label( right_frame, justify="left", anchor="w", text=( "1. 多边形: 逐点点击，点回起点附近可闭合\n" "2. 框选: 拖拽生成矩形，自动写入裁剪框\n" "3. 输入支持:\n" " rect:10,20,200,300\n" " polygon:10,20,100,20,100,100\n" " 纯 4 个数默认矩形，6 个及以上默认多边形\n" "4. 批量回显支持 result 1: (x1 y1 x2 y2) 0.98, label" ), ) self.help_text.pack(fill="x") def _bind_events(self): self.root.bind("", self.undo_last_point) def set_status(self, message, color="darkgreen"): self.status_label.config(text=f"状态: {message}", fg=color) def update_canvas_size(self, width, height): self.canvas.config(scrollregion=(0, 0, width, height)) self.canvas.config(width=min(width, 1100), height=min(height, 760)) def display_image(self): self.canvas.delete("all") self.clear_canvas_state() if not self.image: return self.update_canvas_size(self.image.width, self.image.height) self.photo = ImageTk.PhotoImage(self.image) self.canvas.create_image(0, 0, image=self.photo, anchor="nw") self.redraw_current_shape() def clear_canvas_state(self): self.lines = [] self.point_labels = [] self.box_shape_id = None self.box_text_id = None self.box_preview = None self.box_preview_text = None self.clear_restored_shapes() def clear_restored_shapes(self): for item in self.restored_points + self.restored_lines + self.restored_text_labels + self.restored_rectangles: self.canvas.delete(item) self.restored_points = [] self.restored_lines = [] self.restored_text_labels = [] self.restored_rectangles = [] def update_input_boxes(self): self.coords_text.delete("1.0", tk.END) if self.current_shape: self.coords_text.insert("1.0", self.serialize_annotation(self.current_shape)) self.crop_entry.delete(0, tk.END) if self.box_coords: self.crop_entry.insert(0, f"{self.box_coords[0]},{self.box_coords[1]},{self.box_coords[2]},{self.box_coords[3]}") def redraw_current_shape(self): if not self.current_shape: self.update_input_boxes() return if self.current_shape["type"] == "polygon": self.points = [tuple(point) for point in self.current_shape["points"]] self.draw_polygon(self.points, color="red", editable=True) elif self.current_shape["type"] == "rect": self.box_coords = tuple(self.current_shape["coords"]) self.draw_rect(self.box_coords, color="blue", editable=True) self.update_input_boxes() def draw_polygon(self, points, color="red", editable=False): if not points: return for index, (x, y) in enumerate(points, start=1): oval = self.canvas.create_oval(x - 5, y - 5, x + 5, y + 5, fill=color, outline=color) label = self.canvas.create_text( x + 10, y, text=str(index), fill="green" if editable else color, font=("Arial", 14, "bold"), ) if editable: self.point_labels.extend([oval, label]) else: self.restored_points.append(oval) self.restored_text_labels.append(label) for start, end in zip(points, points[1:]): line = self.canvas.create_line(start, end, fill=color, width=2) if editable: self.lines.append(line) else: self.restored_lines.append(line) if len(points) > 2: line = self.canvas.create_line(points[-1], points[0], fill=color, width=2) if editable: self.lines.append(line) else: self.restored_lines.append(line) def draw_rect(self, coords, color="blue", editable=False, label=None): x1, y1, x2, y2 = coords rect = self.canvas.create_rectangle(x1, y1, x2, y2, outline=color, width=2) width = abs(x2 - x1) height = abs(y2 - y1) text = self.canvas.create_text( min(x1, x2) + 10, min(y1, y2) + 12, anchor="nw", text=label or f"{width}x{height}", fill=color, font=("Arial", 12, "bold"), ) if editable: self.box_shape_id = rect self.box_text_id = text else: self.restored_rectangles.append(rect) self.restored_text_labels.append(text) def enable_polygon_mode(self): self.mode = "polygon" self.canvas.unbind("") self.canvas.unbind("") self.canvas.unbind("") self.canvas.bind("", self.on_click) self.set_status("已切换到多边形模式") def enable_box_mode(self): self.mode = "box" self.canvas.unbind("") self.canvas.bind("", self.on_box_press) self.canvas.bind("", self.on_box_drag) self.canvas.bind("", self.on_box_release) self.set_status("已切换到框选模式") def on_click(self, event): if self.mode != "polygon" or not self.image: return x, y = self.clamp_to_image(event.x, event.y) if self.box_coords: self.box_coords = None if len(self.points) > 2: first_x, first_y = self.points[0] distance = ((x - first_x) ** 2 + (y - first_y) ** 2) ** 0.5 if distance < 20: self.current_shape = {"type": "polygon", "points": self.points[:]} self.display_image() self.set_status("多边形已闭合") return self.points.append((x, y)) self.current_shape = {"type": "polygon", "points": self.points[:]} self.display_image() def on_box_press(self, event): if self.mode != "box" or not self.image: return self.box_start = self.clamp_to_image(event.x, event.y) self.box_coords = None self.current_shape = None self.display_image() def on_box_drag(self, event): if self.mode != "box" or not self.image or not self.box_start: return x0, y0 = self.box_start x1, y1 = self.clamp_to_image(event.x, event.y) if self.box_preview: self.canvas.delete(self.box_preview) if self.box_preview_text: self.canvas.delete(self.box_preview_text) self.box_preview = self.canvas.create_rectangle(x0, y0, x1, y1, outline="blue", width=2, dash=(4, 2)) self.box_preview_text = self.canvas.create_text( min(x0, x1) + 10, min(y0, y1) + 12, anchor="nw", text=f"{abs(x1 - x0)}x{abs(y1 - y0)}", fill="blue", font=("Arial", 12, "bold"), ) def on_box_release(self, event): if self.mode != "box" or not self.image or not self.box_start: return x0, y0 = self.box_start x1, y1 = self.clamp_to_image(event.x, event.y) self.box_start = None self.box_coords = (min(x0, x1), min(y0, y1), max(x0, x1), max(y0, y1)) self.current_shape = {"type": "rect", "coords": self.box_coords} self.display_image() self.set_status("框选完成") def clamp_to_image(self, x, y): if not self.image: return x, y x = min(max(int(self.canvas.canvasx(x)), 0), max(self.image.width - 1, 0)) y = min(max(int(self.canvas.canvasy(y)), 0), max(self.image.height - 1, 0)) return x, y def serialize_annotation(self, annotation): if annotation["type"] == "rect": x1, y1, x2, y2 = annotation["coords"] return f"rect:{x1},{y1},{x2},{y2}" points = annotation["points"] flat = ",".join(f"{x},{y}" for x, y in points) return f"polygon:{flat}" def parse_annotation(self, text): raw = text.strip() if not raw: return None lowered = raw.lower() if lowered.startswith("rect:"): coords = self.parse_ints(raw.split(":", 1)[1]) if len(coords) != 4: raise ValueError("矩形需要 4 个数字") x1, y1, x2, y2 = coords return {"type": "rect", "coords": (min(x1, x2), min(y1, y2), max(x1, x2), max(y1, y2))} if lowered.startswith("polygon:"): coords = self.parse_ints(raw.split(":", 1)[1]) if len(coords) < 6 or len(coords) % 2 != 0: raise ValueError("多边形至少需要 3 个点") points = [(coords[i], coords[i + 1]) for i in range(0, len(coords), 2)] return {"type": "polygon", "points": points} coords = self.parse_ints(raw) if len(coords) == 4: x1, y1, x2, y2 = coords return {"type": "rect", "coords": (min(x1, x2), min(y1, y2), max(x1, x2), max(y1, y2))} if len(coords) >= 6 and len(coords) % 2 == 0: points = [(coords[i], coords[i + 1]) for i in range(0, len(coords), 2)] return {"type": "polygon", "points": points} raise ValueError("无法识别坐标格式") def parse_ints(self, text): numbers = re.findall(r"-?\d+", text) return [int(num) for num in numbers] def load_annotation(self, annotation, from_input=False): self.current_shape = annotation if annotation["type"] == "polygon": self.points = [tuple(point) for point in annotation["points"]] self.box_coords = None else: self.points = [] self.box_coords = tuple(annotation["coords"]) self.display_image() if from_input: self.set_status("已根据输入恢复标注") def save_coordinates(self): if not self.current_shape: self.set_status("请先绘制或恢复标注", "red") return coords_str = self.serialize_annotation(self.current_shape) self.update_input_boxes() print(coords_str) if self.folder_path and self.image_list: image_name = self.image_list[self.current_image_index] self.coordinates[image_name] = coords_str self.write_coordinates_file() self.highlight_images_with_coordinates() self.set_status(f"已保存 {image_name} 的标注") if self.current_image_index < len(self.image_list) - 1: self.next_image() else: self.set_status("已输出当前标注") def write_coordinates_file(self): if not self.folder_path: return data_file = os.path.join(self.folder_path, "data.txt") with open(data_file, "w", encoding="utf-8") as file: for image_name, coords in self.coordinates.items(): file.write(f"{image_name}: {coords}\n") def reset(self): self.points = [] self.current_shape = None self.box_coords = None self.box_start = None self.display_image() self.update_input_boxes() self.set_status("已重置当前标注") def undo_last_point(self, event=None): if self.current_shape and self.current_shape["type"] == "rect": self.current_shape = None self.box_coords = None self.display_image() self.update_input_boxes() self.set_status("已清除矩形框") return if self.points: self.points.pop() if self.points: self.current_shape = {"type": "polygon", "points": self.points[:]} else: self.current_shape = None self.display_image() self.set_status("已撤回一个点") def upload_file(self): file_types = [("Image/Video files", "*.jpg *.jpeg *.png *.bmp *.mp4 *.avi *.mov")] file_path = filedialog.askopenfilename(filetypes=file_types) if not file_path: return self.file_path = file_path if file_path.lower().endswith((".jpg", ".jpeg", ".png", ".bmp")): self.image = Image.open(file_path).convert("RGB") else: cap = cv2.VideoCapture(file_path) success, frame = cap.read() cap.release() if not success: self.set_status("视频首帧读取失败", "red") return frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) self.image = Image.fromarray(frame) self.current_shape = None self.points = [] self.box_coords = None self.display_image() self.set_status(f"已加载文件: {os.path.basename(file_path)}") def select_folder(self): folder_path = filedialog.askdirectory() if not folder_path: return self.folder_path = folder_path self.refresh_image_list() self.load_existing_coordinates() if self.image_list: self.current_image_index = 0 self.load_image_from_folder() self.set_status(f"已选择文件夹: {folder_path}") def refresh_image_list(self): if not self.folder_path: return self.image_list = sorted( [ name for name in os.listdir(self.folder_path) if name.lower().endswith((".jpg", ".jpeg", ".png", ".bmp")) ] ) self.image_listbox.delete(0, tk.END) for image_name in self.image_list: self.image_listbox.insert(tk.END, image_name) self.highlight_images_with_coordinates() def load_existing_coordinates(self): self.coordinates = {} if not self.folder_path: return data_file = os.path.join(self.folder_path, "data.txt") if not os.path.exists(data_file): return with open(data_file, "r", encoding="utf-8") as file: for raw_line in file: line = raw_line.strip() if not line or ": " not in line: continue image_name, coords_str = line.split(": ", 1) self.coordinates[image_name] = coords_str def load_image_from_folder(self): if not self.folder_path or not self.image_list: return image_name = self.image_list[self.current_image_index] image_path = os.path.join(self.folder_path, image_name) self.image = Image.open(image_path).convert("RGB") self.current_shape = None self.points = [] self.box_coords = None self.display_image() self.highlight_current_image() self.restore_coordinates_from_file() self.set_status(f"已加载: {image_name}") def on_image_select(self, event): if not self.image_listbox.curselection(): return self.current_image_index = self.image_listbox.curselection()[0] self.load_image_from_folder() def highlight_current_image(self): for index, image_name in enumerate(self.image_list): color = "blue" if image_name in self.coordinates else "black" self.image_listbox.itemconfig(index, {"fg": color}) if self.image_list: self.image_listbox.itemconfig(self.current_image_index, {"fg": "red"}) self.image_listbox.selection_clear(0, tk.END) self.image_listbox.selection_set(self.current_image_index) self.image_listbox.see(self.current_image_index) def highlight_images_with_coordinates(self): for index, image_name in enumerate(self.image_list): color = "blue" if image_name in self.coordinates else "black" self.image_listbox.itemconfig(index, {"fg": color}) def prev_image(self): if not self.image_list: return self.current_image_index = max(0, self.current_image_index - 1) self.load_image_from_folder() def next_image(self): if not self.image_list: return self.current_image_index = min(len(self.image_list) - 1, self.current_image_index + 1) self.load_image_from_folder() def search_image(self, event=None): keyword = self.search_entry.get().strip().lower() if not keyword: return for index, image_name in enumerate(self.image_list): if keyword in image_name.lower(): self.current_image_index = index self.load_image_from_folder() return self.set_status("未找到匹配图片", "red") def restore_coordinates(self): if not self.image: self.set_status("请先加载图片", "red") return coords_str = self.coords_text.get("1.0", tk.END).strip() if not coords_str: self.set_status("请输入坐标内容", "red") return try: annotation = self.parse_annotation(coords_str) except ValueError as exc: self.set_status(str(exc), "red") return self.load_annotation(annotation, from_input=True) def restore_coordinates_from_file(self): if not self.image_list: return image_name = self.image_list[self.current_image_index] coords_str = self.coordinates.get(image_name) if not coords_str: self.coords_text.delete("1.0", tk.END) self.crop_entry.delete(0, tk.END) return try: annotation = self.parse_annotation(coords_str) except ValueError: self.coords_text.delete("1.0", tk.END) self.coords_text.insert("1.0", coords_str) self.set_status(f"{image_name} 的坐标格式无法自动恢复", "red") return self.load_annotation(annotation) def batch_restore_coordinates(self): if not self.image: self.set_status("请先加载图片", "red") return input_text = self.coords_text.get("1.0", tk.END).strip() if not input_text: self.set_status("请输入批量框内容", "red") return self.clear_restored_shapes() count = 0 pattern = re.compile( r"result\s+(\d+):\s*\(\s*(\d+)\s+(\d+)\s+(\d+)\s+(\d+)\s*\)\s*[\d.]+\s*,\s*([A-Za-z0-9_\-]+)" ) for line in input_text.splitlines(): match = pattern.search(line) if not match: continue result_id, x1, y1, x2, y2, label = match.groups() coords = (int(x1), int(y1), int(x2), int(y2)) self.draw_rect(coords, color="yellow", editable=False, label=f"result {result_id} {label}") count += 1 if count == 0: self.set_status("没有识别到批量框格式", "red") else: self.set_status(f"已回显 {count} 个检测框") def get_crop_box(self): if self.box_coords: return self.box_coords raw = self.crop_entry.get().strip() if not raw: raise ValueError("请先框选或输入裁剪坐标") annotation = self.parse_annotation(raw) if annotation["type"] == "rect": return annotation["coords"] xs = [point[0] for point in annotation["points"]] ys = [point[1] for point in annotation["points"]] return min(xs), min(ys), max(xs), max(ys) def crop_image(self): if not self.image: self.set_status("请先加载图片", "red") return try: x1, y1, x2, y2 = self.get_crop_box() except ValueError as exc: self.set_status(str(exc), "red") return x1 = max(0, min(x1, self.image.width)) y1 = max(0, min(y1, self.image.height)) x2 = max(0, min(x2, self.image.width)) y2 = max(0, min(y2, self.image.height)) if x2 <= x1 or y2 <= y1: self.set_status("裁剪区域无效", "red") return cropped = self.image.crop((x1, y1, x2, y2)) target_size = 640 ratio = min(target_size / cropped.width, target_size / cropped.height) new_width = max(1, int(cropped.width * ratio)) new_height = max(1, int(cropped.height * ratio)) resized = cropped.resize((new_width, new_height), Image.Resampling.LANCZOS) background = Image.new("RGB", (target_size, target_size), (0, 0, 0)) offset = ((target_size - new_width) // 2, (target_size - new_height) // 2) background.paste(resized, offset) save_path = filedialog.asksaveasfilename( title="保存裁剪结果", defaultextension=".jpg", filetypes=[("JPEG files", "*.jpg"), ("PNG files", "*.png")], ) if save_path: background.save(save_path) self.image = background self.current_shape = None self.points = [] self.box_coords = None self.display_image() self.set_status(f"裁剪完成: ({x1},{y1})-({x2},{y2})") def compress_image(self): if not self.image: self.set_status("请先加载图片", "red") return self.image = self.make_640_image(self.image) self.current_shape = None self.points = [] self.box_coords = None self.display_image() self.set_status("当前图片已压缩到 640x640") def compress_all_images(self): if not self.folder_path or not self.image_list: self.set_status("请先选择工作文件夹", "red") return for image_name in self.image_list: image_path = os.path.join(self.folder_path, image_name) image = Image.open(image_path).convert("RGB") compressed = self.make_640_image(image) compressed.save(image_path) self.load_image_from_folder() self.set_status("批量压缩完成") def make_640_image(self, image): target_size = 640 ratio = min(target_size / image.width, target_size / image.height) new_width = max(1, int(image.width * ratio)) new_height = max(1, int(image.height * ratio)) resized = image.resize((new_width, new_height), Image.Resampling.LANCZOS) background = Image.new("RGB", (target_size, target_size), (0, 0, 0)) offset = ((target_size - new_width) // 2, (target_size - new_height) // 2) background.paste(resized, offset) return background def connect_rtsp(self): rtsp_url = simpledialog.askstring("RTSP 连接", "请输入 RTSP 地址:") if not rtsp_url: return self.rtsp_url = rtsp_url self.set_status("正在连接 RTSP...", "blue") self.rtsp_thread = threading.Thread(target=self._connect_rtsp_thread, daemon=True) self.rtsp_thread.start() def _connect_rtsp_thread(self): try: cap = cv2.VideoCapture(self.rtsp_url) cap.set(cv2.CAP_PROP_OPEN_TIMEOUT_MSEC, 5000) if not cap.isOpened(): self.root.after(0, lambda: self.set_status("RTSP 连接失败", "red")) return success, _ = cap.read() if not success: cap.release() self.root.after(0, lambda: self.set_status("RTSP 读取失败", "red")) return self.cap = cap self.root.after(0, lambda: self.set_status("RTSP 已连接")) except Exception as exc: self.root.after(0, lambda: self.set_status(f"RTSP 错误: {exc}", "red")) def capture_rtsp_frame(self): if not self.cap or not self.cap.isOpened(): self.set_status("RTSP 尚未连接", "red") return success, frame = self.cap.read() if not success: self.set_status("RTSP 截图失败", "red") return frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) self.image = Image.fromarray(frame_rgb) self.current_shape = None self.points = [] self.box_coords = None self.display_image() save_path = filedialog.asksaveasfilename( title="保存 RTSP 截图", initialfile=f"rtsp_capture_{time.strftime('%Y%m%d_%H%M%S')}.jpg", defaultextension=".jpg", filetypes=[("JPEG files", "*.jpg"), ("PNG files", "*.png")], ) if save_path: cv2.imwrite(save_path, frame) if self.folder_path: copied_path = os.path.join(self.folder_path, os.path.basename(save_path)) shutil.copy2(save_path, copied_path) self.refresh_image_list() self.set_status("RTSP 截图完成") def disconnect_rtsp(self): if self.cap: self.cap.release() self.cap = None self.set_status("RTSP 已断开") def extract_video_frames(self): video_path = filedialog.askopenfilename( title="选择视频文件", filetypes=[("Video files", "*.mp4 *.avi *.mov *.mkv")], ) if not video_path: return output_folder = filedialog.askdirectory(title="选择输出文件夹") if not output_folder: return target_fps = simpledialog.askinteger("抽帧", "请输入目标帧率(FPS):", initialvalue=5, minvalue=1, maxvalue=60) if not target_fps: return video = cv2.VideoCapture(video_path) original_fps = video.get(cv2.CAP_PROP_FPS) if not original_fps: video.release() self.set_status("无法读取视频帧率", "red") return interval = max(1, int(original_fps / target_fps)) frame_index = 0 saved_count = 0 while True: success, frame = video.read() if not success: break if frame_index % interval == 0: save_path = os.path.join(output_folder, f"frame_{saved_count:05d}.jpg") cv2.imwrite(save_path, frame) saved_count += 1 frame_index += 1 video.release() self.set_status(f"抽帧完成，保存 {saved_count} 张") def collect_labeled_images(self): label_folder = filedialog.askdirectory(title="选择标签文件夹") if not label_folder: return image_source_folder = filedialog.askdirectory(title="选择图片源文件夹") if not image_source_folder: return output_folder = filedialog.askdirectory(title="选择输出文件夹") if not output_folder: return images_out = os.path.join(output_folder, "images") labels_out = os.path.join(output_folder, "labels") os.makedirs(images_out, exist_ok=True) os.makedirs(labels_out, exist_ok=True) count = 0 for filename in os.listdir(label_folder): if not filename.endswith(".txt") or filename == "classes.txt": continue base_name = os.path.splitext(filename)[0] src_label = os.path.join(label_folder, filename) found_image = None for suffix in (".jpg", ".jpeg", ".png", ".bmp"): candidate = os.path.join(image_source_folder, base_name + suffix) if os.path.exists(candidate): found_image = candidate break if not found_image: continue shutil.copy2(found_image, os.path.join(images_out, os.path.basename(found_image))) shutil.copy2(src_label, os.path.join(labels_out, filename)) count += 1 classes_path = os.path.join(label_folder, "classes.txt") if os.path.exists(classes_path): shutil.copy2(classes_path, os.path.join(output_folder, "classes.txt")) self.set_status(f"整理完成，共复制 {count} 组图像和标签") messagebox.showinfo("完成", f"已整理 {count} 组标注数据") def main(): root = tk.Tk() UniversalAnnotationTool(root) root.mainloop() if __name__ == "__main__": main()