| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151 |
- architecture: SSD
- use_gpu: true
- max_iters: 400000
- snapshot_iter: 10000
- log_iter: 20
- metric: COCO
- pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/VGG16_caffe_pretrained.tar
- save_dir: output
- weights: output/ssd_vgg16_512/model_final
- num_classes: 81
- SSD:
- backbone: VGG
- multi_box_head: MultiBoxHead
- output_decoder:
- background_label: 0
- keep_top_k: 200
- nms_eta: 1.0
- nms_threshold: 0.45
- nms_top_k: 400
- score_threshold: 0.01
- VGG:
- depth: 16
- with_extra_blocks: true
- normalizations: [20., -1, -1, -1, -1, -1, -1]
- extra_block_filters: [[256, 512, 1, 2, 3], [128, 256, 1, 2, 3], [128, 256, 1, 2, 3], [128, 256, 1, 2, 3], [128, 256, 1, 1, 4]]
- MultiBoxHead:
- base_size: 512
- aspect_ratios: [[2.], [2., 3.], [2., 3.], [2., 3.], [2., 3.], [2.], [2.]]
- min_ratio: 15
- max_ratio: 90
- min_sizes: [20.0, 51.0, 133.0, 215.0, 296.0, 378.0, 460.0]
- max_sizes: [51.0, 133.0, 215.0, 296.0, 378.0, 460.0, 542.0]
- steps: [8, 16, 32, 64, 128, 256, 512]
- offset: 0.5
- flip: true
- kernel_size: 3
- pad: 1
- LearningRate:
- base_lr: 0.001
- schedulers:
- - !PiecewiseDecay
- gamma: 0.1
- milestones: [280000, 360000]
- - !LinearWarmup
- start_factor: 0.3333333333333333
- steps: 500
- OptimizerBuilder:
- optimizer:
- momentum: 0.9
- type: Momentum
- regularizer:
- factor: 0.0005
- type: L2
- TrainReader:
- inputs_def:
- image_shape: [3, 512, 512]
- fields: ['image', 'gt_bbox', 'gt_class']
- dataset:
- !COCODataSet
- image_dir: train2017
- anno_path: annotations/instances_train2017.json
- dataset_dir: dataset/coco
- sample_transforms:
- - !DecodeImage
- to_rgb: true
- with_mixup: false
- - !RandomDistort
- brightness_lower: 0.875
- brightness_upper: 1.125
- is_order: true
- - !RandomExpand
- fill_value: [104, 117, 123]
- - !RandomCrop
- allow_no_crop: true
- - !NormalizeBox {}
- - !ResizeImage
- interp: 1
- target_size: 512
- use_cv2: false
- - !RandomFlipImage
- is_normalized: true
- - !Permute
- to_bgr: false
- - !NormalizeImage
- is_scale: false
- mean: [104, 117, 123]
- std: [1, 1, 1]
- batch_size: 8
- shuffle: true
- worker_num: 8
- bufsize: 16
- use_process: true
- EvalReader:
- inputs_def:
- image_shape: [3,512,512]
- fields: ['image', 'gt_bbox', 'gt_class', 'im_shape', 'im_id']
- dataset:
- !COCODataSet
- image_dir: val2017
- anno_path: annotations/instances_val2017.json
- dataset_dir: dataset/coco
- sample_transforms:
- - !DecodeImage
- to_rgb: true
- with_mixup: false
- - !ResizeImage
- interp: 1
- target_size: 512
- use_cv2: false
- - !Permute
- to_bgr: false
- - !NormalizeImage
- is_scale: false
- mean: [104, 117, 123]
- std: [1, 1, 1]
- batch_size: 8
- worker_num: 8
- bufsize: 16
- drop_empty: false
- TestReader:
- inputs_def:
- image_shape: [3,512,512]
- fields: ['image', 'im_id', 'im_shape']
- dataset:
- !ImageFolder
- anno_path: annotations/instances_val2017.json
- sample_transforms:
- - !DecodeImage
- to_rgb: true
- with_mixup: false
- - !ResizeImage
- interp: 1
- max_size: 0
- target_size: 512
- use_cv2: true
- - !Permute
- to_bgr: false
- - !NormalizeImage
- is_scale: false
- mean: [104, 117, 123]
- std: [1, 1, 1]
- batch_size: 1
|
