Hi again, moving my previous question to a new thread as it is faster RCNN specific.
In addition, I have one more question on training. I get a lot of “No GT bboxes found in …” in spite of moving images without annotations out of the training folder. I have checked and rechecked the corresponding labels/image_xx.txt file, and the annotations do exist.
No GT bboxes found in image/workspace/nvidia-tlt/data/KITTI/train/images/yyOdbcpCQ6s__000289.jpg
No GT bboxes found in image/workspace/nvidia-tlt/data/KITTI/train/images/YMgJt9yBLgY__001842.jpg
No GT bboxes found in image/workspace/nvidia-tlt/data/KITTI/train/images/vt0hblIsHiY__000120.jpg
No GT bboxes found in image/workspace/nvidia-tlt/data/KITTI/train/images/F2Bw4OLZHq8__001746.jpg
No GT bboxes found in image/workspace/nvidia-tlt/data/KITTI/train/images/vlo00n25e18__000709.jpg
No GT bboxes found in image/workspace/nvidia-tlt/data/KITTI/train/images/6hT5xOszncI__000889.jpg
338/15736 […] - ETA: 2:36:47 - rpn_cls: 0.2377 - rpn_regr: 0.0102 - detector_cls: 0.0515 - detector_regr: 0.0169
No GT bboxes found in image/workspace/nvidia-tlt/data/KITTI/train/images/SLL5ziDWc6k__001346.jpg
339/15736 […] - ETA: 2:36:40 - rpn_cls: 0.2373 - rpn_regr: 0.0102 - detector_cls: 0.0515 - detector_regr: 0.0169
No positive ROIs.
In spite of this, the model trains ok (losses go down) and so I tried to evaluate after a couple of epochs. What I get is:
2020-01-18 11:35:06,570 [INFO] /usr/local/lib/python2.7/dist-packages/iva/faster_rcnn/scripts/test.pyc: 640/642
2020-01-18 11:35:06,742 [INFO] /usr/local/lib/python2.7/dist-packages/iva/faster_rcnn/scripts/test.pyc: Elapsed time = 0.172304868698
2020-01-18 11:35:06,750 [INFO] /usr/local/lib/python2.7/dist-packages/iva/faster_rcnn/scripts/test.pyc: 641/642
2020-01-18 11:35:06,930 [INFO] /usr/local/lib/python2.7/dist-packages/iva/faster_rcnn/scripts/test.pyc: Elapsed time = 0.179259061813
Class AP precision recall
my_class 0.0000 0.0000 0.0000
mAP = 0.0000
Here’s my training spec file:
random_seed: 42
enc_key: 'my_key'
verbose: True
network_config {
input_image_config {
image_type: RGB
image_channel_order: 'bgr'
size_height_width {
height: 576
width: 1024
}
image_channel_mean {
key: 'b'
value: 103.939
}
image_channel_mean {
key: 'g'
value: 116.779
}
image_channel_mean {
key: 'r'
value: 123.68
}
image_scaling_factor: 1.0
}
feature_extractor: "resnet:18"
anchor_box_config {
scale: 64.0
scale: 128.0
scale: 256.0
ratio: 1.0
ratio: 0.5
ratio: 2.0
}
freeze_bn: True
freeze_blocks: 0
freeze_blocks: 1
roi_mini_batch: 256
rpn_stride: 16
conv_bn_share_bias: False
roi_pooling_config {
pool_size: 7
pool_size_2x: False
}
all_projections: True
use_pooling:False
}
training_config {
kitti_data_config {
images_dir: '/workspace/nvidia-tlt/data/KITTI/train/images'
labels_dir: '/workspace/nvidia-tlt/data/KITTI/train/labels'
}
training_data_parser: 'raw_kitti'
data_augmentation {
use_augmentation: True
spatial_augmentation {
hflip_probability: 0.0
vflip_probability: 0.0
zoom_min: 1.0
zoom_max: 1.0
translate_max_x: 0
translate_max_y: 0
}
color_augmentation {
color_shift_stddev: 0.0
hue_rotation_max: 0.0
saturation_shift_max: 0.0
contrast_scale_max: 0.0
contrast_center: 0.5
}
}
num_epochs: 6
class_mapping {
key: 'my_class'
value: 0
}
class_mapping {
key: 'background'
value: 1
}
pretrained_weights: "/workspace/nvidia-tlt/data/faster_rcnn/resnet18.h5"
pretrained_model: ""
output_weights: "/workspace/nvidia-tlt/ckpts/my_frcnn.tltw"
output_model: "/workspace/nvidia-tlt/ckpts/my_frcnn.tlt"
rpn_min_overlap: 0.3
rpn_max_overlap: 0.7
classifier_min_overlap: 0.0
classifier_max_overlap: 0.5
gt_as_roi: False
std_scaling: 1.0
classifier_regr_std {
key: 'x'
value: 10.0
}
classifier_regr_std {
key: 'y'
value: 10.0
}
classifier_regr_std {
key: 'w'
value: 5.0
}
classifier_regr_std {
key: 'h'
value: 5.0
}
rpn_mini_batch: 256
rpn_pre_nms_top_N: 12000
rpn_nms_max_boxes: 2000
rpn_nms_overlap_threshold: 0.7
reg_config {
reg_type: 'L2'
weight_decay: 1e-4
}
optimizer {
adam {
lr: 0.00001
beta_1: 0.9
beta_2: 0.999
decay: 0.0
}
}
lr_scheduler {
step {
base_lr: 0.00001
gamma: 1.0
step_size: 30
}
}
lambda_rpn_regr: 1.0
lambda_rpn_class: 1.0
lambda_cls_regr: 1.0
lambda_cls_class: 1.0
inference_config {
images_dir: '/workspace/nvidia-tlt/data/KITTI/valid/images'
model: '/workspace/nvidia-tlt/ckpts/my_frcnn.epoch6.tlt'
detection_image_output_dir: '/workspace/nvidia-tlt/out/frcnn/images'
labels_dump_dir: '/workspace/nvidia-tlt/out/frcnn/labels_inference'
rpn_pre_nms_top_N: 6000
rpn_nms_max_boxes: 300
rpn_nms_overlap_threshold: 0.7
bbox_visualize_threshold: 0.6
classifier_nms_max_boxes: 300
classifier_nms_overlap_threshold: 0.3
}
evaluation_config {
dataset {
images_dir: '/workspace/nvidia-tlt/data/KITTI/valid/images'
labels_dir: '/workspace/nvidia-tlt/data/KITTI/valid/labels'
}
data_parser: 'raw_kitti'
model: '/workspace/nvidia-tlt/ckpts/my_frcnn.epoch6.tlt'
labels_dump_dir: '/workspace/nvidia-tlt/out/frcnn/labels_evaluation'
rpn_pre_nms_top_N: 6000
rpn_nms_max_boxes: 300
rpn_nms_overlap_threshold: 0.7
classifier_nms_max_boxes: 300
classifier_nms_overlap_threshold: 0.3
object_confidence_thres: 0.0001
use_voc07_11point_metric:False
}
}
Thanks for your help!