{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 生成效果抽查"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 40/40 [00:00<00:00, 171.67it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[{\"Q\": \"What color is the reporter's coat during the interview?',\n",
      "  'Options': [\n",
      "      'A. Red.',\n",
      "      'B. Black.',\n",
      "      'C. Blue.',\n",
      "      'D. White.'\n",
      "  ],\n",
      "  'Answer': 'B\"}]\n",
      "[{\"Q\": \"What color is the man's jacket in the video?', 'Options': ['A. Red.', 'B. Blue.', 'C. Green.', 'D. Black.'], 'Answer': 'B\"}]\n",
      "[{\"Q\": \"What is the person's hands manipulating in the video clip?', 'Options': ['A. A small electronic component.', 'B. A large mechanical part.', 'C. A piece of paper.', 'D. A tool from the workspace.'], 'Answer': 'A\"}]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "import json\n",
    "import os\n",
    "import random\n",
    "import shutil\n",
    "from tqdm import tqdm\n",
    "import re\n",
    "\n",
    "def parse(generated_text):\n",
    "    generated_text = generated_text.strip()\n",
    "    if \"```json\" in generated_text:\n",
    "        generated_text = re.sub(r\"^```json\\s*|\\s*```$\", \"\", generated_text.strip())\n",
    "    try:\n",
    "        data = eval(generated_text)\n",
    "    except:\n",
    "        generated_text = generated_text.replace('\\'Q\\': \\'', \"\\\"Q\\\": \\\"\").replace('\\', \\'A\\': \\'', \"\\\", \\\"A\\\": \\\"\").replace('\\'}', \"\\\"}\")\n",
    "        print(generated_text)\n",
    "        data = eval(generated_text)\n",
    "\n",
    "    return data\n",
    "\n",
    "def load_jsonl(path):\n",
    "    datas = []\n",
    "    with open(path, 'r') as file:\n",
    "        for line in file:\n",
    "            data = json.loads(line)\n",
    "            datas.append(data)\n",
    "    return datas\n",
    "\n",
    "\n",
    "results_dir = '/share/minghao/VideoProjects/Sythesis2/ObjectRecognition/Results/object_0_40'\n",
    "file_names = os.listdir(results_dir)\n",
    "all_datas = []\n",
    "for name in file_names:\n",
    "    file_path = os.path.join(results_dir, name)\n",
    "    if os.path.isdir(file_path):\n",
    "        continue\n",
    "    datas = load_jsonl(file_path)\n",
    "    all_datas.extend(datas)\n",
    "\n",
    "total_size = len(all_datas)\n",
    "check_size = total_size//10\n",
    "check_size = 40\n",
    "check_datas = random.sample(all_datas, check_size)\n",
    "observation_dir = '/share/minghao/VideoProjects/Sythesis2/ObjectRecognition/Observations'\n",
    "\n",
    "for data in tqdm(check_datas):\n",
    "    video_id = data['video_id']\n",
    "    generated_text = data['generated_qa']\n",
    "    try:\n",
    "        result = parse(generated_text)\n",
    "    except:\n",
    "        continue\n",
    "    with open(os.path.join(observation_dir, video_id+'.json'), 'w') as file:\n",
    "        json.dump(result, file, indent=4)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2700/2700 [00:00<00:00, 1360151.43it/s]\n"
     ]
    }
   ],
   "source": [
    "import json\n",
    "from tqdm import tqdm\n",
    "\n",
    "path = '/share/minghao/VideoProjects/VideoChat-Flash/lmms-eval_videochat/videochat-flash-7B@448_eval_log_videomme.json'\n",
    "\n",
    "with open(path, 'r') as f:\n",
    "    datas = json.load(f)\n",
    "\n",
    "score_task_type = {}\n",
    "score_duration_task_type = {\n",
    "    'short': {},\n",
    "    'medium': {},\n",
    "    'long': {}\n",
    "}\n",
    "\n",
    "record_error = {\n",
    "    'short': [],\n",
    "    'medium': [],\n",
    "    'long': []\n",
    "}\n",
    "\n",
    "for data in tqdm(datas['logs']):\n",
    "    duration = data['doc']['duration']\n",
    "    task_type = data['doc']['task_type']\n",
    "    if task_type != 'Action Reasoning':\n",
    "        continue\n",
    "    pred_answer = data['videomme_percetion_score']['pred_answer']\n",
    "    answer = data['videomme_percetion_score']['answer']\n",
    "    if pred_answer == answer:\n",
    "        correct_flag = True\n",
    "    else:\n",
    "        correct_flag = False\n",
    "\n",
    "    if not correct_flag:\n",
    "        info = {\n",
    "            'url' : data['doc']['url'],\n",
    "            'Q': data['doc']['question'],\n",
    "            'Options': data['doc']['options'],\n",
    "            'Answer': answer,\n",
    "            'pred': pred_answer\n",
    "        }\n",
    "\n",
    "\n",
    "        record_error[duration].append(info)\n",
    "\n",
    "import os\n",
    "\n",
    "ob_dir = '/share/minghao/VideoProjects/Sythesis/ActionReasoning/ErrorObservation'\n",
    "for duration, error_infos in record_error.items():\n",
    "    ob_path = os.path.join(ob_dir, f'{duration}.json')\n",
    "    with open(ob_path, 'w') as file:\n",
    "        json.dump(error_infos, file, indent=4)\n",
    "    "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}