148 lines
4.7 KiB
Python
148 lines
4.7 KiB
Python
#!/usr/bin/env python
|
||
"""
|
||
分析推理结果 - 不导入mmdet3d模块避免CUDA问题
|
||
"""
|
||
import os
|
||
import sys
|
||
|
||
# 添加项目路径但不导入mmdet3d
|
||
sys.path.insert(0, '/workspace/bevfusion')
|
||
|
||
# 只导入基础库
|
||
import pickle
|
||
import numpy as np
|
||
|
||
def load_results_without_mmdet3d(filepath):
|
||
"""不导入mmdet3d模块的情况下加载结果"""
|
||
# 临时重命名mmdet3d模块避免自动导入
|
||
import sys
|
||
mmdet3d_modules = [m for m in sys.modules.keys() if m.startswith('mmdet3d')]
|
||
for mod in mmdet3d_modules:
|
||
del sys.modules[mod]
|
||
|
||
# 阻止mmdet3d的自动导入
|
||
sys.modules['mmdet3d'] = None
|
||
sys.modules['mmdet3d.core'] = None
|
||
|
||
try:
|
||
with open(filepath, 'rb') as f:
|
||
# 使用纯Python方式加载，避免触发__reduce__方法中的导入
|
||
import pickle
|
||
results = pickle.load(f)
|
||
return results
|
||
except Exception as e:
|
||
print(f"标准pickle加载失败: {e}")
|
||
return None
|
||
|
||
def analyze_results(results):
|
||
"""分析推理结果"""
|
||
print("="*60)
|
||
print("BEVFusion 单Batch推理结果分析")
|
||
print("="*60)
|
||
|
||
print(f"总样本数量: {len(results)}")
|
||
|
||
for sample_idx, sample in enumerate(results):
|
||
print(f"\n--- 样本 {sample_idx} ---")
|
||
print(f"数据类型: {type(sample)}")
|
||
print(f"包含字段: {list(sample.keys())}")
|
||
|
||
# 分析BEV分割
|
||
if 'masks_bev' in sample:
|
||
masks = sample['masks_bev']
|
||
print(f"BEV分割形状: {masks.shape}")
|
||
print("BEV分割数据类型:", type(masks))
|
||
|
||
if hasattr(masks, 'shape'): # numpy array or torch tensor
|
||
print(f"BEV分割数据类型: {masks.dtype if hasattr(masks, 'dtype') else 'unknown'}")
|
||
print("每个类别的像素统计:")
|
||
|
||
class_names = ['drivable_area', 'ped_crossing', 'walkway', 'stop_line', 'carpark_area', 'divider']
|
||
for i in range(min(6, masks.shape[0])):
|
||
if i < len(class_names):
|
||
# 计算激活像素数量
|
||
if hasattr(masks, 'cpu'): # torch tensor
|
||
mask_data = masks[i].cpu().numpy()
|
||
else: # numpy array
|
||
mask_data = masks[i]
|
||
|
||
activated_pixels = (mask_data > 0.5).sum()
|
||
total_pixels = mask_data.size
|
||
percentage = activated_pixels / total_pixels * 100
|
||
|
||
print("15s")
|
||
|
||
# 分析Ground Truth
|
||
if 'gt_masks_bev' in sample:
|
||
gt_masks = sample['gt_masks_bev']
|
||
print(f"Ground Truth形状: {gt_masks.shape}")
|
||
|
||
# 分析3D检测
|
||
if 'boxes_3d' in sample:
|
||
boxes = sample['boxes_3d']
|
||
print(f"3D检测框数量: {len(boxes)}")
|
||
|
||
if 'scores_3d' in sample:
|
||
scores = sample['scores_3d']
|
||
if len(scores) > 0:
|
||
if hasattr(scores, 'cpu'):
|
||
scores_np = scores.cpu().numpy()
|
||
else:
|
||
scores_np = np.array(scores)
|
||
|
||
print(".3f")
|
||
print(".3f")
|
||
|
||
if 'labels_3d' in sample:
|
||
labels = sample['labels_3d']
|
||
if len(labels) > 0:
|
||
if hasattr(labels, 'cpu'):
|
||
labels_np = labels.cpu().numpy()
|
||
else:
|
||
labels_np = np.array(labels)
|
||
|
||
unique_labels, counts = np.unique(labels_np, return_counts=True)
|
||
print("检测类别分布:")
|
||
for label, count in zip(unique_labels, counts):
|
||
print(f" 类别 {int(label)}: {count} 个")
|
||
|
||
print("-" * 30)
|
||
|
||
def main():
|
||
# 推理结果文件路径
|
||
result_file = '/data/infer_test/20251120_124755/one_batch_results.pkl'
|
||
|
||
if not os.path.exists(result_file):
|
||
print(f"结果文件不存在: {result_file}")
|
||
# 查找可能的其他结果文件
|
||
import glob
|
||
possible_files = glob.glob('/data/infer_test/*/one_batch_results.pkl')
|
||
if possible_files:
|
||
print("找到以下可能的结果文件:")
|
||
for f in possible_files:
|
||
print(f" {f}")
|
||
result_file = possible_files[-1] # 使用最新的
|
||
print(f"使用: {result_file}")
|
||
else:
|
||
return
|
||
|
||
print(f"分析结果文件: {result_file}")
|
||
print(f"文件大小: {os.path.getsize(result_file)} bytes")
|
||
|
||
# 尝试加载结果
|
||
results = load_results_without_mmdet3d(result_file)
|
||
|
||
if results is None:
|
||
print("无法加载结果文件")
|
||
return
|
||
|
||
# 分析结果
|
||
analyze_results(results)
|
||
|
||
print("\n" + "="*60)
|
||
print("分析完成")
|
||
print("="*60)
|
||
|
||
if __name__ == '__main__':
|
||
main()
|