Update bevformer train network

Chapter07-课程展望与总结/README.md

@@ -2,8 +2,8 @@
  * @Author: Charmve [email protected]
  * @Date: 2023-10-10 10:49:13
  * @LastEditors: Charmve [email protected]
- * @LastEditTime: 2023-10-10 10:53:07
- * @FilePath: /OccNet-Course/Chapter06-课程展望与总结/README.md
+ * @LastEditTime: 2024-01-22 19:55:01
+ * @FilePath: /OccNet-Course/Chapter07-课程展望与总结/README.md
  * @Version: 1.0.1
  * @Blogs: charmve.blog.csdn.net
  * @GitHub: https://github.com/Charmve
@@ -13,16 +13,27 @@
  * Licensed under the MIT License.
 -->
 
-在本专题课程的课程展望和总结中，主要从OCC与Nerf的关系、端到端的实现、大模型三个方面做未来展望，以及对本课程做一个总结。
+在本专题课程的课程展望和总结中，主要从算法框架、数据、仿真和其他四个方面做未来展望，以及对本课程做一个总结。
 
-1. OCC与Nerf的关系
+- <h4>算法框架</h4>
+  - 数据驱动的端到端 UniAD
+  - 大模型 LMDrive [关于大模型和自动驾驶的几个迷思](关于大模型和自动驾驶的几个迷思.md)
+  - 世界模型：Drive-WM、DriveDreamer
+  - 矢量地图在线建图：MapTRv2
+  - BEV-OCC-Transformer: OccFormer、OccWorld、Occupancy Flow
 
-- []()
+- <h4>数据</h4>
+  - 4D数据自动标注: 
+    - OCC与Nerf联合标注
+    - [面向BEV感知的4D标注方案](https://zhuanlan.zhihu.com/p/642735557?utm_psn=1706841959639998464)
+  - 数据生成：DrivingDiffusion、[MagicDrive](https://zhuanlan.zhihu.com/p/675303127)、UrbanSyn
 
-2. 端到端的实现
+- <h4>仿真</h4>
+  - UniSim
+  - DRIVE Sim
 
-- []()
+- <h4>其他</h4>
+  - 舱驾一体
+  - AI 编译器: MLIR、、XLA、Triton
+  - 模型剪枝、模型蒸馏、模型压缩、模型量化（PTQ、QAT）
 
-3. 自动驾驶大模型
-
-- [关于大模型和自动驾驶的几个迷思](关于大模型和自动驾驶的几个迷思.md)

code/BEVFormer/docs/prepare_dataset.md

@@ -15,6 +15,7 @@ unzip can_bus.zip
 
 *We genetate custom annotation files which are different from mmdet3d's*
 ```
+export PYTHONPATH=$PYTHONPATH:./
 python tools/create_data.py nuscenes --root-path ./data/nuscenes --out-dir ./data/nuscenes --extra-tag nuscenes --version v1.0 --canbus ./data
 ```
 

code/BEVFormer/projects/configs/bevformer/bevformer_base.py

@@ -2,7 +2,7 @@
     '../datasets/custom_nus-3d.py',
     '../_base_/default_runtime.py'
 ]
-#
+
 plugin = True
 plugin_dir = 'projects/mmdet3d_plugin/'
 
@@ -11,8 +11,6 @@
 point_cloud_range = [-51.2, -51.2, -5.0, 51.2, 51.2, 3.0]
 voxel_size = [0.2, 0.2, 8]
 
-
-
 img_norm_cfg = dict(
     mean=[103.530, 116.280, 123.675], std=[1.0, 1.0, 1.0], to_rgb=False)
 # For nuScenes we usually do 10-class detection

code/BEVFormer/projects/configs/bevformer/bevformer_small.py

@@ -1,10 +1,10 @@
 # BEvFormer-small consumes at lease 10500M GPU memory
 # compared to bevformer_base, bevformer_small has
-# smaller BEV: 200*200 -> 150*150
-# less encoder layers: 6 -> 3
-# smaller input size: 1600*900 -> (1600*900)*0.8
-# multi-scale feautres -> single scale features (C5)
-# with_cp of backbone = True
+#  - smaller BEV: 200*200 -> 150*150
+#  - less encoder layers: 6 -> 3
+#  - smaller input size: 1600*900 -> (1600*900)*0.8
+#  - multi-scale feautres -> single scale features (C5)
+#  - with_cp of backbone = True
 
 _base_ = [
     '../datasets/custom_nus-3d.py',

code/BEVFormer/projects/configs/bevformer/bevformer_tiny.py

@@ -1,10 +1,10 @@
 # BEvFormer-tiny consumes at lease 6700M GPU memory
 # compared to bevformer_base, bevformer_tiny has
-# smaller backbone: R101-DCN -> R50
-# smaller BEV: 200*200 -> 50*50
-# less encoder layers: 6 -> 3
-# smaller input size: 1600*900 -> 800*450
-# multi-scale feautres -> single scale features (C5)
+#  - smaller backbone: R101-DCN -> R50
+#  - smaller BEV: 200*200 -> 50*50
+#  - less encoder layers: 6 -> 3
+#  - smaller input size: 1600*900 -> 800*450
+#  - multi-scale feautres -> single scale features (C5)
 
 
 _base_ = [

code/BEVFormer/tests/bevformer_base/20240122/pts_bbox/results_nusc.json

@@ -0,0 +1,5 @@
+{
+    "sample": [
+        {"configurations": "test1"}
+    ]
+}

code/BEVFormer/tools/analysis_tools/visual.py

@@ -4,45 +4,32 @@
 # ---------------------------------------------
 
 import mmcv
-from nuscenes.nuscenes import NuScenes
+
 from PIL import Image
-from nuscenes.utils.geometry_utils import view_points, box_in_image, BoxVisibility, transform_matrix
 from typing import Tuple, List, Iterable
-import matplotlib.pyplot as plt
+
 import numpy as np
-from PIL import Image
-from matplotlib import rcParams
-from matplotlib.axes import Axes
-from pyquaternion import Quaternion
-from PIL import Image
 from matplotlib import rcParams
+import matplotlib.pyplot as plt
 from matplotlib.axes import Axes
 from pyquaternion import Quaternion
 from tqdm import tqdm
+
+from nuscenes.nuscenes import NuScenes
 from nuscenes.utils.data_classes import LidarPointCloud, RadarPointCloud, Box
 from nuscenes.utils.geometry_utils import view_points, box_in_image, BoxVisibility, transform_matrix
 from nuscenes.eval.common.data_classes import EvalBoxes, EvalBox
 from nuscenes.eval.detection.data_classes import DetectionBox
 from nuscenes.eval.detection.utils import category_to_detection_name
 from nuscenes.eval.detection.render import visualize_sample
 
-
-
-
 cams = ['CAM_FRONT',
  'CAM_FRONT_RIGHT',
  'CAM_BACK_RIGHT',
  'CAM_BACK',
  'CAM_BACK_LEFT',
  'CAM_FRONT_LEFT']
 
-import numpy as np
-import matplotlib.pyplot as plt
-from nuscenes.utils.data_classes import LidarPointCloud, RadarPointCloud, Box
-from PIL import Image
-from matplotlib import rcParams
-
-
 def render_annotation(
         anntoken: str,
         margin: float = 10,
@@ -84,6 +71,7 @@ def render_annotation(
     fig, axes = plt.subplots(1, num_cam + 1, figsize=(18, 9))
     select_cams = [sample_record['data'][cam] for cam in select_cams]
     print('bbox in cams:', select_cams)
+
     # Plot LIDAR view.
     lidar = sample_record['data']['LIDAR_TOP']
     data_path, boxes, camera_intrinsic = nusc.get_sample_data(lidar, selected_anntokens=[anntoken])
@@ -142,8 +130,6 @@ def render_annotation(
     if out_path is not None:
         plt.savefig(out_path)
 
-
-
 def get_sample_data(sample_data_token: str,
                     box_vis_level: BoxVisibility = BoxVisibility.ANY,
                     selected_anntokens=None,
@@ -205,8 +191,6 @@ def get_sample_data(sample_data_token: str,
 
     return data_path, box_list, cam_intrinsic
 
-
-
 def get_predicted_data(sample_data_token: str,
                        box_vis_level: BoxVisibility = BoxVisibility.ANY,
                        selected_anntokens=None,
@@ -269,9 +253,6 @@ def get_predicted_data(sample_data_token: str,
 
     return data_path, box_list, cam_intrinsic
 
-
-
-
 def lidiar_render(sample_token, data,out_path=None):
     bbox_gt_list = []
     bbox_pred_list = []
@@ -316,7 +297,6 @@ def lidiar_render(sample_token, data,out_path=None):
     print('blue is the predited result')
     visualize_sample(nusc, sample_token, gt_annotations, pred_annotations, savepath=out_path+'_bev')
 
-
 def get_color(category_name: str):
     """
     Provides the default colors based on the category names.
@@ -347,7 +327,6 @@ def get_color(category_name: str):
             return nusc.colormap[key]
     return [0, 0, 0]
 
-
 def render_sample_data(
         sample_toekn: str,
         with_anns: bool = True,
@@ -471,7 +450,7 @@ def render_sample_data(
 if __name__ == '__main__':
     nusc = NuScenes(version='v1.0-trainval', dataroot='./data/nuscenes', verbose=True)
     # render_annotation('7603b030b42a4b1caa8c443ccc1a7d52')
-    bevformer_results = mmcv.load('test/bevformer_base/Thu_Jun__9_16_22_37_2022/pts_bbox/results_nusc.json')
+    bevformer_results = mmcv.load('tests/bevformer_base/20240122/pts_bbox/results_nusc.json')
     sample_token_list = list(bevformer_results['results'].keys())
     for id in range(0, 10):
         render_sample_data(sample_token_list[id], pred_data=bevformer_results, out_path=sample_token_list[id])

code/BEVFusion/docker/notebook

@@ -2,7 +2,7 @@
 echo $BASE_URL
 
 if [ `whoami` = "root" ];then
-		# 启动jupyter并设置对应路径
+	# 启动jupyter并设置对应路径
     /opt/conda/bin/jupyter lab --ip=0.0.0.0 --LabApp.base_url=$BASE_URL --allow-root
 else
     /opt/conda/bin/jupyter lab --ip=0.0.0.0 --LabApp.base_url=$BASE_URL

tools/app/config.json

@@ -1,6 +1,6 @@
 {
-    "name": "Qbot VIPs",
-    "url": "ufund-me.github.io/Qbot",
+    "name": "OccCourse VIPs",
+    "url": "maiweiai.site",
     "author": "Charmve",
     "isNonProfit": true,
     "address": {

tools/app/myapp.py

@@ -24,7 +24,7 @@
 # mail_license = os.getenv("MAIL_LICENSE")
 mail_license = "szfiwfywaakhieda"
 # 收件人邮箱，可以为多个收件人
-mail_receivers = ["[email protected]", "zhangwei@qcraft.ai"]
+mail_receivers = ["[email protected]", "zhangwei@maiwei.ai"]
 
 
 @app.route('/')

tools/app/pull_issues.py

@@ -18,8 +18,8 @@
 def pull_issues(data = data):
     # 发送 POST 请求
     # https://github.com/Charmve/100days
-    repo_owner = "UFund-Me"
-    repo_name = "Qbot"
+    repo_owner = "Charmve"
+    repo_name = "OccNet-Course"
     url = f"https://api.github.com/repos/{repo_owner}/{repo_name}/issues"
     response = re.post(url, headers=headers, json=data)
 
@@ -52,7 +52,7 @@ def close_github_issues(issue_url):
 
 if __name__ == '__main__':
     issues_data = {
-        "title": "🌈 {today} 来自Qbot的今日AI选股推送",
+        "title": "💡 {today} 来自OccCource更新提醒",
         "body": "{content}"
     }
     pull_issues(issues_data)