humanSeg for iOS

[zhangjunhst]

问题确认 Search before asking

• [X] 我已经搜索过问题，但是没有找到解答。I have searched the question and found no related answer.

请提出你的问题 Please ask your question

iOS端 使用paddle-lite代码替换模型后返回的数据与效果不符合，相同的模型安卓根据 https://github.com/PaddlePaddle/PaddleSeg/issues/279#issuecomment-638821411 修改后返回正确，而iOS不知道改如何做对应的改造

[zhangjunhst]

Model: DeepLabv3p Transforms:

• ResizeByLong: {long_size: 512}
• Padding: im_padding_value: [127.5, 127.5, 127.5] label_padding_value: 255 target_size: 512
• Normalize: max_val: [255.0, 255.0, 255.0] mean: [0.5, 0.5, 0.5] min_val: [0, 0, 0] std: [0.5, 0.5, 0.5] TransformsMode: BGR _Attributes: eval_metrics: {miou: 0.6704429236012107} fixed_input_shape: null labels: [background, pointer, scale] model_type: segmenter num_classes: 3 _ModelInputsOutputs: test_inputs:
  • [image, image] test_outputs:
  • [pred, unsqueeze2_0.tmp_0]
  • [logit, softmax_0.tmp_0] _init_params: {aspp_with_sep_conv: true, backbone: MobileNetV3_large_x1_0_ssld, class_weight: null, decoder_use_sep_conv: true, enable_decoder: true, encoder_with_aspp: true, ignore_index: 255, input_channel: 3, num_classes: 3, output_stride: 16, pooling_crop_size: null, use_bce_loss: false, use_dice_loss: false} completed_epochs: 0 status: Infer version: 1.2.8

模型的参数