ValueError: Dimension 1 in both shapes must be equal, but are 13 and 14.

Hello,

I am new to this and coming across an issue that I do not quite understand. Appreciate any help here!

I have input images of size 800x1066 (sometimes 1067). I've generated my tfrecords and beginning to train (run model_main_tf2.py). I am using Tensorflow Object Detection API, and made some changes to the default EfficientNet-b0 configuration file.

The issue: I am getting a ValueError: Exception encountered when calling layer "1_dn_lvl_6/combine" (type BiFPNCombineLayer).

In particular: ValueError: Dimension 1 in both shapes must be equal, but are 13 and 14. Shapes are [4,13,17,64] an d [4,14,18,64]. From merging shape 0 with other shapes. for '{{node EfficientDet-D0/bifpn/node_02/1_dn_lvl_ 6/combine/stack}} = Pack[N=2, T=DT_FLOAT, axis=-1](EfficientDet-D0/bifpn/node_00/0_up_lvl_6/input_0_up_lvl_5/downsa mple_max_x2/MaxPool, EfficientDet-D0/bifpn/node_02/1_dn_lvl_6/input_0_up_lvl_7/nearest_neighbor_upsampling_x2/neare st_neighbor_upsampling/Reshape)' with input shapes: [4,13,17,64], [4,14,18,64].

I do not quite understand what this error means. It seems like it is related to my input images. Is it incorrect to be using input images that are rectangular (in my case 800x1067)?

This is what my pipeline.config looks like:

# SSD with EfficientNet-b0 + BiFPN feature extractor,
# shared box predictor and focal loss (a.k.a EfficientDet-d0).
# See EfficientDet, Tan et al, https://arxiv.org/abs/1911.09070
# See Lin et al, https://arxiv.org/abs/1708.02002
# Trained on COCO, initialized from an EfficientNet-b0 checkpoint.
#
# Train on TPU-8

model {
  ssd {
    inplace_batchnorm_update: true
    freeze_batchnorm: false
    num_classes: 1
    add_background_class: false
    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
        use_matmul_gather: true
      }
    }
    similarity_calculator {
      iou_similarity {
      }
    }
    encode_background_as_zeros: true
    anchor_generator {
      multiscale_anchor_generator {
        min_level: 3
        max_level: 7
        anchor_scale: 4.0
        aspect_ratios: [1.0, 2.0, 0.5]
        scales_per_octave: 3
      }
    }
    image_resizer {
      fixed_shape_resizer {
        height: 800
        width: 1067
        resize_method: AREA
      }
    }
    box_predictor {
      weight_shared_convolutional_box_predictor {
        depth: 64
        class_prediction_bias_init: -4.6
        conv_hyperparams {
          force_use_bias: true
          activation: SWISH
          regularizer {
            l2_regularizer {
              weight: 0.00004
            }
          }
          initializer {
            random_normal_initializer {
              stddev: 0.01
              mean: 0.0
            }
          }
          batch_norm {
            scale: true
            decay: 0.99
            epsilon: 0.001
          }
        }
        num_layers_before_predictor: 3
        kernel_size: 3
        use_depthwise: true
      }
    }
    feature_extractor {
      type: 'ssd_efficientnet-b0_bifpn_keras'
      bifpn {
        min_level: 3
        max_level: 7
        num_iterations: 3
        num_filters: 64
      }
      conv_hyperparams {
        force_use_bias: true
        activation: SWISH
        regularizer {
          l2_regularizer {
            weight: 0.00004
          }
        }
        initializer {
          truncated_normal_initializer {
            stddev: 0.03
            mean: 0.0
          }
        }
        batch_norm {
          scale: true,
          decay: 0.99,
          epsilon: 0.001,
        }
      }
    }
    loss {
      classification_loss {
        weighted_sigmoid_focal {
          alpha: 0.25
          gamma: 1.5
        }
      }
      localization_loss {
        weighted_smooth_l1 {
        }
      }
      classification_weight: 1.0
      localization_weight: 1.0
    }
    normalize_loss_by_num_matches: true
    normalize_loc_loss_by_codesize: true
    post_processing {
      batch_non_max_suppression {
        score_threshold: 1e-8
        iou_threshold: 0.5
        max_detections_per_class: 150
        max_total_detections: 150
      }
      score_converter: SIGMOID
    }
  }
}

train_config: {
  batch_size: 4
  sync_replicas: true
  startup_delay_steps: 0
  replicas_to_aggregate: 8
  use_bfloat16: true
  num_steps: 30000
  data_augmentation_options {
    random_horizontal_flip {
    }
  }
  data_augmentation_options {
    random_vertical_flip {
    }
  }
  data_augmentation_options {
    random_adjust_brightness{
      }
  }
  data_augmentation_options {
    random_adjust_contrast{
    }
  }
  data_augmentation_options {
    random_adjust_hue{
    }
  }
  data_augmentation_options {
    random_adjust_saturation{
    }
  }
  optimizer {
    momentum_optimizer: {
      learning_rate: {
        cosine_decay_learning_rate {
          learning_rate_base: 8e-2
          total_steps: 30000
          warmup_learning_rate: .001
          warmup_steps: 250
        }
      }
      momentum_optimizer_value: 0.9
    }
    use_moving_average: false
  }
  max_number_of_boxes: 150
  unpad_groundtruth_tensors: false
}

train_input_reader: {
  label_map_path: "<<my input path>>"
  tf_record_input_reader {
    input_path: "<<my input path>>"
  }
}

eval_config: {
  metrics_set: "coco_detection_metrics"
  use_moving_averages: false
  batch_size: 1;
}

eval_input_reader: {
  label_map_path: "<<my input path>>"
  shuffle: false
  num_epochs: 1
  tf_record_input_reader {
    input_path: "<<my input path>>"
  }
}

6. System information

• OS Platform and Distribution: macOS 12.0
• TensorFlow version (use command below): v2.7.0
• Python version: 3.7.12