Table 1 Details of H-RT-DETR and the compared methods
From: Plant recognition of maize seedling stage in UAV remote sensing images based on H-RT-DETR
Methods | Backbone | Optimizer | Pre-trained | Hyperparameters for training |
|---|---|---|---|---|
YOLOv5 | CSPDarknet53 | SGD | False | Learning rate:initial was 0.001, min was 0.00001,delay type was cos |
YOLOv7 | CSPDarknet53 | SGD | False | Learning rate:initial was 0.001, min was 0.00001,delay type was cos |
YOLOv8 | CSPDarknet53 | SGD | False | Learning rate:initial was 0.001, min was 0.00001,delay type was cos |
YOLOX | CSPDarknet53 | SGD | False | Learning rate:initial was 0.001, min was 0.00001,delay type was cos |
RT-DETR | PResNet | AdamW | False | Encoder:HybridEncoder(in_channels:[512, 1024, 2048],feat_strides:[8, 16, 32], hidden_dim = 256,nhead = 8,activation function ='gelu') Decoder:RTDETRTransformer(feat_channels: [256, 256, 256],feat_strides: [8, 16, 32], hidden_dim = 256,num_levels = 3,num_queries = 300,num_decoder_layers: 6,num_denoising: 100) |
H-RT-DETR | HFR | AdamW | False | Encoder:HybridEncoder(in_channels:[512, 1024, 2048],feat_strides:[8, 16, 32], hidden_dim = 256,nhead = 8,activation function ='gelu') Decoder:RTDETRTransformer(feat_channels: [256, 256, 256],feat_strides: [8, 16, 32], hidden_dim = 256,num_levels = 3,num_queries = 300,num_decoder_layers: 6,num_denoising: 100) |