|
| 1 | +model { |
| 2 | + ssd { |
| 3 | + num_classes: 4 |
| 4 | + image_resizer { |
| 5 | + keep_aspect_ratio_resizer { |
| 6 | + min_dimension: 512 |
| 7 | + max_dimension: 512 |
| 8 | + pad_to_max_dimension: true |
| 9 | + } |
| 10 | + } |
| 11 | + feature_extractor { |
| 12 | + type: "ssd_efficientnet-b0_bifpn_keras" |
| 13 | + conv_hyperparams { |
| 14 | + regularizer { |
| 15 | + l2_regularizer { |
| 16 | + weight: 4e-05 |
| 17 | + } |
| 18 | + } |
| 19 | + initializer { |
| 20 | + truncated_normal_initializer { |
| 21 | + mean: 0.0 |
| 22 | + stddev: 0.03 |
| 23 | + } |
| 24 | + } |
| 25 | + activation: SWISH |
| 26 | + batch_norm { |
| 27 | + decay: 0.99 |
| 28 | + scale: true |
| 29 | + epsilon: 0.001 |
| 30 | + } |
| 31 | + force_use_bias: true |
| 32 | + } |
| 33 | + bifpn { |
| 34 | + min_level: 3 |
| 35 | + max_level: 7 |
| 36 | + num_iterations: 3 |
| 37 | + num_filters: 64 |
| 38 | + } |
| 39 | + } |
| 40 | + box_coder { |
| 41 | + faster_rcnn_box_coder { |
| 42 | + y_scale: 10.0 |
| 43 | + x_scale: 10.0 |
| 44 | + height_scale: 5.0 |
| 45 | + width_scale: 5.0 |
| 46 | + } |
| 47 | + } |
| 48 | + matcher { |
| 49 | + argmax_matcher { |
| 50 | + matched_threshold: 0.5 |
| 51 | + unmatched_threshold: 0.5 |
| 52 | + ignore_thresholds: false |
| 53 | + negatives_lower_than_unmatched: true |
| 54 | + force_match_for_each_row: true |
| 55 | + use_matmul_gather: true |
| 56 | + } |
| 57 | + } |
| 58 | + similarity_calculator { |
| 59 | + iou_similarity { |
| 60 | + } |
| 61 | + } |
| 62 | + box_predictor { |
| 63 | + weight_shared_convolutional_box_predictor { |
| 64 | + conv_hyperparams { |
| 65 | + regularizer { |
| 66 | + l2_regularizer { |
| 67 | + weight: 4e-05 |
| 68 | + } |
| 69 | + } |
| 70 | + initializer { |
| 71 | + random_normal_initializer { |
| 72 | + mean: 0.0 |
| 73 | + stddev: 0.01 |
| 74 | + } |
| 75 | + } |
| 76 | + activation: SWISH |
| 77 | + batch_norm { |
| 78 | + decay: 0.99 |
| 79 | + scale: true |
| 80 | + epsilon: 0.001 |
| 81 | + } |
| 82 | + force_use_bias: true |
| 83 | + } |
| 84 | + depth: 64 |
| 85 | + num_layers_before_predictor: 3 |
| 86 | + kernel_size: 3 |
| 87 | + class_prediction_bias_init: -4.6 |
| 88 | + use_depthwise: true |
| 89 | + } |
| 90 | + } |
| 91 | + anchor_generator { |
| 92 | + multiscale_anchor_generator { |
| 93 | + min_level: 3 |
| 94 | + max_level: 7 |
| 95 | + anchor_scale: 4.0 |
| 96 | + aspect_ratios: 1.0 |
| 97 | + aspect_ratios: 2.0 |
| 98 | + aspect_ratios: 0.5 |
| 99 | + scales_per_octave: 3 |
| 100 | + } |
| 101 | + } |
| 102 | + post_processing { |
| 103 | + batch_non_max_suppression { |
| 104 | + score_threshold: 1e-08 |
| 105 | + iou_threshold: 0.5 |
| 106 | + max_detections_per_class: 100 |
| 107 | + max_total_detections: 100 |
| 108 | + } |
| 109 | + score_converter: SIGMOID |
| 110 | + } |
| 111 | + normalize_loss_by_num_matches: true |
| 112 | + loss { |
| 113 | + localization_loss { |
| 114 | + weighted_smooth_l1 { |
| 115 | + } |
| 116 | + } |
| 117 | + classification_loss { |
| 118 | + weighted_sigmoid_focal { |
| 119 | + gamma: 1.5 |
| 120 | + alpha: 0.25 |
| 121 | + } |
| 122 | + } |
| 123 | + classification_weight: 1.0 |
| 124 | + localization_weight: 1.0 |
| 125 | + } |
| 126 | + encode_background_as_zeros: true |
| 127 | + normalize_loc_loss_by_codesize: true |
| 128 | + inplace_batchnorm_update: true |
| 129 | + freeze_batchnorm: false |
| 130 | + add_background_class: false |
| 131 | + } |
| 132 | +} |
| 133 | +train_config { |
| 134 | + batch_size: 16 |
| 135 | + data_augmentation_options { |
| 136 | + random_horizontal_flip { |
| 137 | + } |
| 138 | + } |
| 139 | + data_augmentation_options { |
| 140 | + random_scale_crop_and_pad_to_square { |
| 141 | + output_size: 512 |
| 142 | + scale_min: 0.1 |
| 143 | + scale_max: 2.0 |
| 144 | + } |
| 145 | + } |
| 146 | + sync_replicas: true |
| 147 | + optimizer { |
| 148 | + momentum_optimizer { |
| 149 | + learning_rate { |
| 150 | + cosine_decay_learning_rate { |
| 151 | + learning_rate_base: 0.08 |
| 152 | + total_steps: 300000 |
| 153 | + warmup_learning_rate: 0.001 |
| 154 | + warmup_steps: 2500 |
| 155 | + } |
| 156 | + } |
| 157 | + momentum_optimizer_value: 0.9 |
| 158 | + } |
| 159 | + use_moving_average: false |
| 160 | + } |
| 161 | + fine_tune_checkpoint: "efficientdet_d0_coco17_tpu-32/checkpoint/ckpt-0" |
| 162 | + num_steps: 4000 |
| 163 | + startup_delay_steps: 0.0 |
| 164 | + replicas_to_aggregate: 8 |
| 165 | + max_number_of_boxes: 100 |
| 166 | + unpad_groundtruth_tensors: false |
| 167 | + fine_tune_checkpoint_type: "detection" |
| 168 | + use_bfloat16: true |
| 169 | + fine_tune_checkpoint_version: V2 |
| 170 | +} |
| 171 | +train_input_reader { |
| 172 | + label_map_path: "/content/labelmap.pbtxt" |
| 173 | + tf_record_input_reader { |
| 174 | + input_path: "/content/train.record" |
| 175 | + } |
| 176 | +} |
| 177 | +eval_config { |
| 178 | + metrics_set: "coco_detection_metrics" |
| 179 | + use_moving_averages: false |
| 180 | + batch_size: 16 |
| 181 | +} |
| 182 | +eval_input_reader { |
| 183 | + label_map_path: "/content/labelmap.pbtxt" |
| 184 | + shuffle: false |
| 185 | + num_epochs: 1 |
| 186 | + tf_record_input_reader { |
| 187 | + input_path: "/content/test.record" |
| 188 | + } |
| 189 | +} |
0 commit comments