config.json

{
  "backbone": {
    "embed_dim": 256,
    "embed_layer": "ConvPyramidPatchEmbed",
    "ffn_layer": "mlp",
    "ffn_ratio": 4,
    "interaction_indexes": [
      10,
      11
    ],
    "name": "ecvittplus",
    "num_heads": 4,
    "num_levels": 3,
    "patch_size": 16,
    "proj_dim": null,
    "skip_load_backbone": false
  },
  "decoder": {
    "activation": "silu",
    "aux_loss": true,
    "box_noise_scale": 1.0,
    "cross_attn_method": "default",
    "dim_feedforward": 1024,
    "dropout": 0.0,
    "eps": 0.01,
    "eval_idx": -1,
    "eval_spatial_size": [
      640,
      640
    ],
    "feat_channels": [
      256,
      256,
      256
    ],
    "feat_strides": [
      8,
      16,
      32
    ],
    "hidden_dim": 256,
    "label_noise_ratio": 0.5,
    "layer_scale": 1,
    "learn_query_content": false,
    "mask_downsample_ratio": null,
    "nhead": 8,
    "num_classes": 80,
    "num_denoising": 100,
    "num_layers": 4,
    "num_levels": 3,
    "num_points": [
      3,
      6,
      3
    ],
    "num_queries": 300,
    "query_select_method": "default",
    "reg_max": 32,
    "reg_scale": 4,
    "share_bbox_head": false,
    "share_score_head": false
  },
  "encoder": {
    "act": "silu",
    "csp_type": "csp2",
    "depth_mult": 0.67,
    "dim_feedforward": 512,
    "dropout": 0.0,
    "eval_spatial_size": null,
    "expansion": 0.75,
    "feat_strides": [
      8,
      16,
      32
    ],
    "fuse_op": "sum",
    "hidden_dim": 256,
    "in_channels": [
      256,
      256,
      256
    ],
    "nhead": 8,
    "num_encoder_layers": 1,
    "pe_temperature": 10000,
    "use_encoder_idx": [
      2
    ]
  },
  "postprocessor": {
    "num_classes": 80,
    "num_top_queries": 300,
    "remap_mscoco_category": false,
    "use_focal_loss": true
  }
}