我正在尝试使用Tensorflow Object Detection API(Mask RCNN)训练实例分割模型,并遵循here的说明。
我正在使用预训练的mask_rcnn_resnet50_atrous_coco
来初始化权重,并为该模型调整this样本配置文件。我根据create_coco_tf_record.py创建了带有训练和评估集掩码的tfrecord文件。
我能够成功运行训练脚本,但问题是,除GPU内存外,它占用45GB
的{{1}}左右。除此之外,一切运行良好,我能够完成高达10k步的训练,之后它决定需要更多RAM并占用RAM
左右,这会导致我的系统崩溃。当我在训练后运行评估脚本时,会发生同样的事情。以下是我的系统规格:
当我在GPU上运行模型时,我不确定为什么tensorflow需要这么多RAM。我只有60GB
和1 foreground class
个培训样本,每张图片最多500
个。
以下是来自50 objects/masks
和System Monitor
以及我的nvidia-smi
文件的一些屏幕截图。如果需要,我还会上传我的脚本来创建tfrecord文件。
这是我的pipeline_config:
pipeline_config
更新1:
我尝试在# Mask R-CNN with Resnet-50 (v1), Atrous version
# Configured for MSCOCO Dataset.
# Users should configure the fine_tune_checkpoint field in the train config as
# well as the label_map_path and input_path fields in the train_input_reader and
# eval_input_reader. Search for "PATH_TO_BE_CONFIGURED" to find the fields that
# should be configured.
model {
faster_rcnn {
num_classes: 1
image_resizer {
keep_aspect_ratio_resizer {
min_dimension: 300
max_dimension: 400
}
}
number_of_stages: 3
feature_extractor {
type: 'faster_rcnn_resnet50'
first_stage_features_stride: 8
}
first_stage_anchor_generator {
grid_anchor_generator {
scales: [0.25, 0.5, 1.0, 2.0]
aspect_ratios: [0.5, 1.0, 2.0]
height_stride: 8
width_stride: 8
}
}
first_stage_atrous_rate: 2
first_stage_box_predictor_conv_hyperparams {
op: CONV
regularizer {
l2_regularizer {
weight: 0.0
}
}
initializer {
truncated_normal_initializer {
stddev: 0.01
}
}
}
first_stage_nms_score_threshold: 0.0
first_stage_nms_iou_threshold: 0.7
first_stage_max_proposals: 300
first_stage_localization_loss_weight: 2.0
first_stage_objectness_loss_weight: 1.0
initial_crop_size: 14
maxpool_kernel_size: 2
maxpool_stride: 2
second_stage_box_predictor {
mask_rcnn_box_predictor {
use_dropout: true
dropout_keep_probability: 0.5
predict_instance_masks: true
mask_height: 33
mask_width: 33
mask_prediction_conv_depth: 0
mask_prediction_num_conv_layers: 4
fc_hyperparams {
op: FC
regularizer {
l2_regularizer {
weight: 0.0
}
}
initializer {
variance_scaling_initializer {
factor: 1.0
uniform: true
mode: FAN_AVG
}
}
}
conv_hyperparams {
op: CONV
regularizer {
l2_regularizer {
weight: 0.0
}
}
initializer {
truncated_normal_initializer {
stddev: 0.01
}
}
}
}
}
second_stage_post_processing {
batch_non_max_suppression {
score_threshold: 0.0
iou_threshold: 0.6
max_detections_per_class: 100
max_total_detections: 300
}
score_converter: SOFTMAX
}
second_stage_localization_loss_weight: 2.0
second_stage_classification_loss_weight: 1.0
second_stage_mask_prediction_loss_weight: 4.0
second_stage_batch_size: 4
}
}
train_config: {
batch_size: 1
optimizer {
momentum_optimizer: {
learning_rate: {
manual_step_learning_rate {
initial_learning_rate: 0.0003
schedule {
step: 0
learning_rate: .0003
}
schedule {
step: 900000
learning_rate: .00003
}
schedule {
step: 1200000
learning_rate: .000003
}
}
}
momentum_optimizer_value: 0.9
}
use_moving_average: false
}
gradient_clipping_by_norm: 10.0
fine_tune_checkpoint: "/media/ahmed/1A6E52446E5218B9/Projects/TF/MaskRCNN/pretrained_models/mask_rcnn_resnet50_atrous_coco_2018_01_28/model.ckpt"
from_detection_checkpoint: true
# Note: The below line limits the training process to 200K steps, which we
# empirically found to be sufficient enough to train the pets dataset. This
# effectively bypasses the learning rate schedule (the learning rate will
# never decay). Remove the below line to train indefinitely.
num_steps: 50000
data_augmentation_options {
random_horizontal_flip {
}
}
}
train_input_reader: {
tf_record_input_reader {
input_path: "/media/ahmed/1A6E52446E5218B9/Projects/TF/MaskRCNN/train_mask.record"
}
label_map_path: "/media/ahmed/1A6E52446E5218B9/Projects/TF/label_map.pbtxt"
load_instance_masks: true
mask_type: PNG_MASKS
}
eval_config: {
num_examples: 200
num_visualizations : 200
# Note: The below line limits the evaluation process to 10 evaluations.
# Remove the below line to evaluate indefinitely.
max_evals: 10
}
eval_input_reader: {
tf_record_input_reader {
input_path: "/media/ahmed/1A6E52446E5218B9/Projects/TF/MaskRCNN/val_mask.record"
}
label_map_path: "/media/ahmed/1A6E52446E5218B9/Projects/TF/label_map.pbtxt"
load_instance_masks: true
mask_type: PNG_MASKS
shuffle: false
num_readers: 1
}
上运行object_detection_tutorial.ipynb。只要我不调用掩码输出节点,它就可以运行2 GB的GPU内存。我得到正确的盒子,课程和分数。当我尝试在frozen_inference_graph
中获得detection_masks:0
的输出时,我的8GB GPU内存不足。我尝试在CPU模式下运行脚本,发现我得到了正确的掩码,我的RAM内存使用率从未超过8GB(从2.5GB)。在运行sess.run()
和train.py
脚本时,同样的模型占用了我的整个GPU以及45GB的RAM。