在训练ssd mobilenet时，我在每个评估点都得到了这一点：
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000
Tensorboard不会像地面真实注释那样在右侧图像上显示检测（边界框）。即使创建了tf记录，我也可以确认注释已被解析。我修改了create_tf_record.py来记录边界框：
I1220 15:44:48.170442 139795475470144 create_coco_tf_record.py:222] Found groundtruth annotations. Building annotations index.
I1220 15:44:48.170788 139795475470144 create_coco_tf_record.py:236] 0 images are missing annotations.
I1220 15:44:48.170870 139795475470144 create_coco_tf_record.py:241] On image 0 of 60
I1220 15:44:48.172753 139795475470144 create_coco_tf_record.py:137] [0.043229 0.083333 0.963889 0.998148]
I1220 15:44:48.172867 139795475470144 create_coco_tf_record.py:137] [0.329167 0.363021 0.922222 0.963889]
...
I1220 15:44:48.342363 139795475470144 create_coco_tf_record.py:249] Finished writing, skipped 0 annotations.
我尝试过不同版本的tensorflow，1.12、1.14、1.13。我尝试了不同版本的ssd / mobilenet，甚至还屏蔽了rcnn。我已经验证了json文件，据我了解，它们已正确转换为记录。
这是我的管道：
model {
ssd {
num_classes: 6
box_coder {
faster_rcnn_box_coder {
y_scale: 10.0
x_scale: 10.0
height_scale: 5.0
width_scale: 5.0
}
}
matcher {
argmax_matcher {
matched_threshold: 0.5
unmatched_threshold: 0.5
ignore_thresholds: false
negatives_lower_than_unmatched: true
force_match_for_each_row: true
}
}
similarity_calculator {
iou_similarity {
}
}
anchor_generator {
ssd_anchor_generator {
num_layers: 6
min_scale: 0.1
max_scale: 0.3
aspect_ratios: 1.0
aspect_ratios: 2.0
aspect_ratios: 0.5
aspect_ratios: 3.0
aspect_ratios: 0.3333
}
}
image_resizer {
fixed_shape_resizer {
height: 300
width: 300
}
}
box_predictor {
convolutional_box_predictor {
min_depth: 0
max_depth: 0
num_layers_before_predictor: 0
use_dropout: true
dropout_keep_probability: 0.5
kernel_size: 1
box_code_size: 4
apply_sigmoid_to_scores: false
conv_hyperparams {
activation: RELU_6,
regularizer {
l2_regularizer {
weight: 0.00004
}
}
initializer {
truncated_normal_initializer {
stddev: 0.03
mean: 0.0
}
}
batch_norm {
train: true,
scale: true,
center: true,
decay: 0.9997,
epsilon: 0.001,
}
}
}
}
feature_extractor {
type: 'ssd_mobilenet_v1'
min_depth: 16
depth_multiplier: 1.0
conv_hyperparams {
activation: RELU_6,
regularizer {
l2_regularizer {
weight: 0.00004
}
}
initializer {
truncated_normal_initializer {
stddev: 0.03
mean: 0.0
}
}
batch_norm {
train: true,
scale: true,
center: true,
decay: 0.9997,
epsilon: 0.001,
}
}
}
loss {
classification_loss {
weighted_sigmoid {
}
}
localization_loss {
weighted_smooth_l1 {
}
}
hard_example_miner {
num_hard_examples: 3000
iou_threshold: 0.99
loss_type: CLASSIFICATION
max_negatives_per_positive: 3
min_negatives_per_image: 0
}
classification_weight: 1.0
localization_weight: 1.0
}
normalize_loss_by_num_matches: true
post_processing {
batch_non_max_suppression {
score_threshold: 1e-8
iou_threshold: 0.6
max_detections_per_class: 50
max_total_detections: 50
}
score_converter: SIGMOID
}
}
}
train_config: {
batch_size: 32
optimizer {
rms_prop_optimizer: {
learning_rate: {
exponential_decay_learning_rate {
initial_learning_rate: 0.004
decay_steps: 800720
decay_factor: 0.95
}
}
momentum_optimizer_value: 0.9
decay: 0.9
epsilon: 1.0
}
}
fine_tune_checkpoint: "/home/ssd/model.ckpt"
from_detection_checkpoint: true
load_all_detection_checkpoint_vars: true
data_augmentation_options {
random_horizontal_flip {
}
}
data_augmentation_options {
ssd_random_crop {
}
}
}
train_input_reader: {
tf_record_input_reader {
input_path: "/home/ssd/coco_train.record-00000-of-00001"
}
label_map_path: "/home/ssd/label_map.pbtxt"
}
eval_config: {
metrics_set: "coco_detection_metrics"
num_examples: 1100
}
eval_input_reader: {
tf_record_input_reader {
input_path: "/home/ssd/coco_val.record-00000-of-00001"
}
label_map_path: "/home/ssd/label_map.pbtxt"
shuffle: false
num_readers: 1
}
起初，我认为问题出在我的注释上，但是我已经使用COCO_Viewer来测试注释，并且可以正确查看它们。我还使用了自己的脚本，可以解析json并将边界框覆盖在图像源上。
这是我train.json https://pastebin.com/ymHS00iN
的粘贴框我在培训ssd / mobilenet时从未遇到过如此麻烦，并且想看看是否有人遇到了这个问题，因为我完全感到困惑。任何想法/评论都欢迎：）
Tensorboard example, not displaying bounding boxes on ground truth data (right-side)
