From 2f5831a593507d74d8f4ee6f4f8170b0a8a695e3 Mon Sep 17 00:00:00 2001 From: kidd <13522042+best-kidd24@users.noreply.github.com> Date: Mon, 13 Nov 2017 18:27:09 +0800 Subject: [PATCH] change TwoImageData to ImageSegData --- .../config/embs/solver_voc_train_aug.prototxt | 28 ++ scripts/segaware/config/embs/test.prototxt | 10 +- .../config/embs/test_voc_val.prototxt | 32 +- scripts/segaware/config/embs/train.prototxt | 10 +- .../config/embs/train_voc_train_aug.prototxt | 330 ++++++++++++++++++ .../config/vgg/solver_voc_train_aug.prototxt | 4 +- scripts/segaware/config/vgg/test.prototxt | 16 +- .../config/vgg/test_voc_test.prototxt | 38 +- .../segaware/config/vgg/test_voc_val.prototxt | 36 +- scripts/segaware/config/vgg/train.prototxt | 20 +- .../config/vgg/train_voc_train_aug.prototxt | 26 +- 11 files changed, 472 insertions(+), 78 deletions(-) create mode 100644 scripts/segaware/config/embs/solver_voc_train_aug.prototxt create mode 100644 scripts/segaware/config/embs/train_voc_train_aug.prototxt diff --git a/scripts/segaware/config/embs/solver_voc_train_aug.prototxt b/scripts/segaware/config/embs/solver_voc_train_aug.prototxt new file mode 100644 index 0000000..b73480e --- /dev/null +++ b/scripts/segaware/config/embs/solver_voc_train_aug.prototxt @@ -0,0 +1,28 @@ +train_net: "/mnt/data1/kidd/segaware/segaware/config/embs/train_voc_train_aug.prototxt" +type: "Adam" + +# these are set by default +#momentum: 0.9 +#momentum2: 0.999 +#delta = 0.00000001 + + + + + + +base_lr: 0.001 + +lr_policy: "fixed" +#gamma: 0.5 +#stepsize: 200000 + +max_iter: 24000 + +display: 10 + +snapshot: 1000 + +snapshot_prefix: "/mnt/data1/kidd/segaware/segaware/model/embs/train_voc_train_aug" + +solver_mode: GPU diff --git a/scripts/segaware/config/embs/test.prototxt b/scripts/segaware/config/embs/test.prototxt index 8fb42a5..58e192c 100644 --- a/scripts/segaware/config/embs/test.prototxt +++ b/scripts/segaware/config/embs/test.prototxt @@ -1,7 +1,7 @@ name: "${NET_ID}" layer { name: "data" - type: "TwoImageData" + type: "ImageSegData" top: "data" top: "label" image_data_param { @@ -10,10 +10,10 @@ layer { batch_size: ${BATCH_SIZE} shuffle: false } - two_image_data_param { - first_is_color: true - second_is_color: false - } +#two_image_data_param { +# first_is_color: true +# second_is_color: false +# } transform_param { mean_value: 104.008 mean_value: 116.669 diff --git a/scripts/segaware/config/embs/test_voc_val.prototxt b/scripts/segaware/config/embs/test_voc_val.prototxt index 1e100cf..a32f776 100644 --- a/scripts/segaware/config/embs/test_voc_val.prototxt +++ b/scripts/segaware/config/embs/test_voc_val.prototxt @@ -1,13 +1,13 @@ -name: "embs_31_L1_coco" +name: "embs" layer { name: "data" type: "TwoImageData" top: "data" top: "label" image_data_param { - root_folder: "/opt/home/aharley/datasets/VOC2012" - source: "segaware/list/voc_val.txt" - batch_size: 1 + root_folder: "/mnt/data1/kidd/segaware/dataset/VOC2012/VOCtrainval" + source: "/mnt/data1/kidd/segaware/segaware/list/voc_val.txt" + batch_size: 15 shuffle: false } two_image_data_param { @@ -18,7 +18,7 @@ layer { mean_value: 104.008 mean_value: 116.669 mean_value: 122.675 - crop_size: 513 + crop_size: 161 mirror: false } } @@ -158,8 +158,8 @@ layer { name: "conv21_resized" top: "conv21_resized" interp_param { - height: 513 - width: 513 + height: 161 + width: 161 } } layer { @@ -168,8 +168,8 @@ layer { name: "conv22_resized" top: "conv22_resized" interp_param { - height: 513 - width: 513 + height: 161 + width: 161 } } layer { @@ -178,8 +178,8 @@ layer { name: "conv31_resized" top: "conv31_resized" interp_param { - height: 513 - width: 513 + height: 161 + width: 161 } } # Concat, and weight @@ -214,7 +214,7 @@ layer { name: "parity" type: "Im2parity" convolution_param { - kernel_size: 5 pad: 6 dilation: 3 + kernel_size: 3 pad: 1 dilation: 1 stride: 1 } } @@ -224,7 +224,7 @@ layer { name: "weighted_avg_dist" type: "Im2dist" convolution_param { - kernel_size: 5 pad: 6 dilation: 3 + kernel_size: 3 pad: 1 dilation: 1 stride: 1 } im2dist_param { @@ -238,7 +238,7 @@ layer { bottom: "parity" top: "loss_avg" convolution_param { - kernel_size: 5 pad: 6 dilation: 3 + kernel_size: 3 pad: 1 dilation: 1 stride: 1 } dist_loss_param { @@ -252,8 +252,8 @@ layer { type: "MatWrite" bottom: "weighted_avg" mat_write_param { - prefix: "segaware/features/embs_31_L1_coco/voc_val/fc8/" - source: "segaware/list/voc_val_id.txt" + prefix: "/mnt/data1/kidd/segaware/segaware/features/embs/voc_val/fc8/" + source: "/mnt/data1/kidd/segaware/segaware/list/voc_val_id.txt" strip: 0 period: 1 } diff --git a/scripts/segaware/config/embs/train.prototxt b/scripts/segaware/config/embs/train.prototxt index 8f9288d..dfe0bf6 100644 --- a/scripts/segaware/config/embs/train.prototxt +++ b/scripts/segaware/config/embs/train.prototxt @@ -1,7 +1,7 @@ name: "${NET_ID}" layer { name: "data" - type: "TwoImageData" + type: "ImageSegData" top: "data" top: "label" image_data_param { @@ -10,10 +10,10 @@ layer { batch_size: ${BATCH_SIZE} shuffle: true } - two_image_data_param { - first_is_color: true - second_is_color: false - } +#two_image_data_param { +# first_is_color: true +# second_is_color: false +# } transform_param { mean_value: 104.008 mean_value: 116.669 diff --git a/scripts/segaware/config/embs/train_voc_train_aug.prototxt b/scripts/segaware/config/embs/train_voc_train_aug.prototxt new file mode 100644 index 0000000..a47eac9 --- /dev/null +++ b/scripts/segaware/config/embs/train_voc_train_aug.prototxt @@ -0,0 +1,330 @@ +name: "embs" +layer { + name: "data" + type: "ImageSegData" + top: "data" + top: "label" + image_data_param { + root_folder: "/mnt/data1/kidd/segaware/dataset/VOC2012/VOCtrainval" + source: "/mnt/data1/kidd/segaware/segaware/list/voc_train_aug.txt" + batch_size: 15 + shuffle: true + } +#two_image_data_param { +# first_is_color: true +# second_is_color: false +# } + transform_param { + mean_value: 104.008 + mean_value: 116.669 + mean_value: 122.675 + crop_size: 161 + mirror: true + } + include: { phase: TRAIN } +} +# rgb for emb +layer { + bottom: "data" + name: "rgb_emb" + top: "rgb_emb" + type: "Power" + power_param { + scale: 0.0039215 # 1/255 + } +} +# Embedding layers +layer { + bottom: "data" + top: "conv1_1" + name: "conv1_1" + type: "Convolution" + param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + convolution_param { + num_output: 64 + kernel_size: 3 pad: 1 + stride: 1 + weight_filler { type: "xavier" } + bias_filler { type: "constant" } + } +} +layer { + bottom: "conv1_1" + top: "conv1_1" + name: "relu1_1" + type: "ReLU" +} +layer { + bottom: "conv1_1" + top: "conv1_2" + name: "conv1_2" + type: "Convolution" + param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + convolution_param { + num_output: 64 + kernel_size: 3 pad: 1 + weight_filler { type: "xavier" } + bias_filler { type: "constant" } + } +} +layer { + bottom: "conv1_2" + top: "conv1_2" + name: "relu1_2" + type: "ReLU" +} +layer { + bottom: "conv1_2" + top: "pool1" + name: "pool1" + type: "Pooling" + pooling_param { + pool: MAX + kernel_size: 3 pad: 1 + stride: 2 + } +} +layer { + bottom: "pool1" + top: "conv2_1" + name: "conv2_1" + type: "Convolution" + param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + convolution_param { + num_output: 128 + kernel_size: 3 pad: 1 + weight_filler { type: "xavier" } + bias_filler { type: "constant" } + } +} +layer { + bottom: "conv2_1" + top: "conv2_1" + name: "relu2_1" + type: "ReLU" +} +layer { + bottom: "conv2_1" + top: "conv2_2" + name: "conv2_2" + type: "Convolution" + param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + convolution_param { + num_output: 128 + kernel_size: 3 pad: 1 + weight_filler { type: "xavier" } + bias_filler { type: "constant" } + } +} +layer { + bottom: "conv2_2" + top: "conv2_2" + name: "relu2_2" + type: "ReLU" +} +layer { + bottom: "conv2_2" + top: "pool2" + name: "pool2" + type: "Pooling" + pooling_param { + pool: MAX + kernel_size: 3 pad: 1 + stride: 2 + } +} +layer { + bottom: "pool2" + top: "conv3_1" + name: "conv3_1" + type: "Convolution" + param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + convolution_param { + num_output: 256 + kernel_size: 3 pad: 1 + weight_filler { type: "xavier" } + bias_filler { type: "constant" } + } +} +layer { + bottom: "conv3_1" + top: "conv3_1" + name: "relu3_1" + type: "ReLU" +} +# resize, for uniformity +layer { + type: "Interp" + bottom: "conv2_1" + name: "conv21_resized" + top: "conv21_resized" + interp_param { + height: 161 + width: 161 + } +} +layer { + type: "Interp" + bottom: "conv2_2" + name: "conv22_resized" + top: "conv22_resized" + interp_param { + height: 161 + width: 161 + } +} +layer { + type: "Interp" + bottom: "conv3_1" + name: "conv31_resized" + top: "conv31_resized" + interp_param { + height: 161 + width: 161 + } +} +# Concat, and weight +layer { + bottom: "rgb_emb" + bottom: "conv1_1" + bottom: "conv1_2" + bottom: "conv21_resized" + bottom: "conv22_resized" + bottom: "conv31_resized" + top: "fused" + name: "fused" + type: "Concat" +} +layer { + bottom: "fused" + top: "weighted_avg" + name: "weighted_avg" + type: "Convolution" + param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + convolution_param { + num_output: 64 + kernel_size: 1 pad: 0 + weight_filler { type: "xavier" } + bias_filler { type: "constant" } + } +} +# Add a narrow loss to the fused thing +layer { + bottom: "label" + top: "parity" + name: "parity" + type: "Im2parity" + convolution_param { + kernel_size: 3 pad: 1 dilation: 1 + stride: 1 + } +} +layer { + bottom: "weighted_avg" + top: "weighted_avg_dist" + name: "weighted_avg_dist" + type: "Im2dist" + convolution_param { + kernel_size: 3 pad: 1 dilation: 1 + stride: 1 + } + im2dist_param { + norm: L1 + } +} +layer { + bottom: "weighted_avg_dist" + bottom: "parity" + top: "loss_avg" + name: "loss_avg" + type: "DistLoss" + convolution_param { + kernel_size: 3 pad: 1 dilation: 1 + stride: 1 + } + dist_loss_param { + alpha: 0.5 + beta: 2 + ignore_label: 255 + } +} +# Add a wide loss to the fused thing +layer { + bottom: "label" + top: "parity2" + name: "parity2" + type: "Im2parity" + convolution_param { + kernel_size: 3 pad: 1 dilation: 2 + stride: 1 + } +} +layer { + bottom: "weighted_avg" + top: "weighted_avg_dist2" + name: "weighted_avg_dist2" + type: "Im2dist" + convolution_param { + kernel_size: 3 pad: 1 dilation: 2 + stride: 1 + } + im2dist_param { + norm: L1 + } +} +layer { + bottom: "weighted_avg_dist2" + bottom: "parity2" + top: "loss_avg2" + name: "loss_avg2" + type: "DistLoss" + convolution_param { + kernel_size: 3 pad: 1 dilation: 2 + stride: 1 + } + dist_loss_param { + alpha: 0.5 + beta: 2 + ignore_label: 255 + } +} +# Add a wider loss to the fused thing +layer { + bottom: "label" + top: "parity3" + name: "parity3" + type: "Im2parity" + convolution_param { + kernel_size: 3 pad: 1 dilation: 5 + stride: 1 + } +} +layer { + bottom: "weighted_avg" + top: "weighted_avg_dist3" + name: "weighted_avg_dist3" + type: "Im2dist" + convolution_param { + kernel_size: 3 pad: 1 dilation: 5 + stride: 1 + } + im2dist_param { + norm: L1 + } +} +layer { + bottom: "weighted_avg_dist3" + bottom: "parity3" + top: "loss_avg3" + name: "loss_avg3" + type: "DistLoss" + convolution_param { + kernel_size: 3 pad: 1 dilation: 5 + stride: 1 + } + dist_loss_param { + alpha: 0.5 + beta: 2 + ignore_label: 255 + } +} diff --git a/scripts/segaware/config/vgg/solver_voc_train_aug.prototxt b/scripts/segaware/config/vgg/solver_voc_train_aug.prototxt index 3c8d974..ca95391 100644 --- a/scripts/segaware/config/vgg/solver_voc_train_aug.prototxt +++ b/scripts/segaware/config/vgg/solver_voc_train_aug.prototxt @@ -1,4 +1,4 @@ -train_net: "segaware/config/segaware-all-largeFOV/train_voc_train_aug.prototxt" +train_net: "/mnt/data1/kidd/segaware/segaware/config/vgg/train_voc_train_aug.prototxt" lr_policy: "step" gamma: 0.1 @@ -11,5 +11,5 @@ momentum: 0.9 weight_decay: 0.0005 snapshot: 60000 -snapshot_prefix: "segaware/model/segaware-all-largeFOV/train_voc_train_aug" +snapshot_prefix: "/mnt/data1/kidd/segaware/segaware/model/vgg/train_voc_train_aug" solver_mode: GPU diff --git a/scripts/segaware/config/vgg/test.prototxt b/scripts/segaware/config/vgg/test.prototxt index 973df40..9de5e30 100644 --- a/scripts/segaware/config/vgg/test.prototxt +++ b/scripts/segaware/config/vgg/test.prototxt @@ -20,19 +20,24 @@ name: "${NET_ID}" layer { name: "data" - type: "TwoImageData" + type: "ImageSegData" top: "data" top: "label" + top: "data_dim" + include { + phase: TEST + } image_data_param { root_folder: "${DATA_ROOT}" source: "${EXP}/list/${TEST_SET}.txt" batch_size: ${BATCH_SIZE} shuffle: false + label_type: NONE } - two_image_data_param { - first_is_color: true - second_is_color: false - } +#two_image_data_param { +# first_is_color: true +# second_is_color: false +# } transform_param { mean_value: 104.008 mean_value: 116.669 @@ -1131,6 +1136,7 @@ layer { } layer { bottom: "label" + bottom: "data_dim" name: "silence" type: "Silence" } diff --git a/scripts/segaware/config/vgg/test_voc_test.prototxt b/scripts/segaware/config/vgg/test_voc_test.prototxt index 57741ba..0ec8cf6 100644 --- a/scripts/segaware/config/vgg/test_voc_test.prototxt +++ b/scripts/segaware/config/vgg/test_voc_test.prototxt @@ -16,19 +16,23 @@ # 161; 81; 41; 21; 11 # -name: "segaware-all-largeFOV" +name: "vgg" layer { name: "data" - type: "ImageData" + type: "TwoImageData" top: "data" top: "label" image_data_param { - root_folder: "/opt/home/aharley/datasets/VOC2012" + root_folder: "/mnt/data1/kidd/segaware/dataset/VOC2012/VOCtest" source: "segaware/list/voc_test.txt" batch_size: 1 shuffle: false } + two_image_data_param { + first_is_color: true + second_is_color: false + } transform_param { mean_value: 104.008 mean_value: 116.669 @@ -311,7 +315,7 @@ layer { top: "conv1_1" name: "conv1_1" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 64 kernel_size: 3 pad: 1 bottom_is_im2col: true @@ -346,7 +350,7 @@ layer { top: "conv1_2" name: "conv1_2" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 64 kernel_size: 3 pad: 1 bottom_is_im2col: true @@ -440,7 +444,7 @@ layer { top: "conv2_1" name: "conv2_1" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 128 kernel_size: 3 pad: 1 bottom_is_im2col: true @@ -474,7 +478,7 @@ layer { top: "conv2_2" name: "conv2_2" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 128 kernel_size: 3 pad: 1 bottom_is_im2col: true @@ -568,7 +572,7 @@ layer { top: "conv3_1" name: "conv3_1" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 256 kernel_size: 3 pad: 1 bottom_is_im2col: true @@ -602,7 +606,7 @@ layer { top: "conv3_2" name: "conv3_2" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 256 kernel_size: 3 pad: 1 bottom_is_im2col: true @@ -636,7 +640,7 @@ layer { top: "conv3_3" name: "conv3_3" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 256 kernel_size: 3 pad: 1 bottom_is_im2col: true @@ -731,7 +735,7 @@ layer { top: "conv4_1" name: "conv4_1" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 512 kernel_size: 3 pad: 1 bottom_is_im2col: true @@ -765,7 +769,7 @@ layer { top: "conv4_2" name: "conv4_2" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 512 kernel_size: 3 pad: 1 bottom_is_im2col: true @@ -799,7 +803,7 @@ layer { top: "conv4_3" name: "conv4_3" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 512 kernel_size: 3 pad: 1 bottom_is_im2col: true @@ -880,7 +884,7 @@ layer { top: "conv5_1" name: "conv5_1" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 512 kernel_size: 3 pad: 2 dilation: 2 bottom_is_im2col: true @@ -914,7 +918,7 @@ layer { top: "conv5_2" name: "conv5_2" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 512 kernel_size: 3 pad: 2 dilation: 2 bottom_is_im2col: true @@ -948,7 +952,7 @@ layer { top: "conv5_3" name: "conv5_3" type: "Convolution" - param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2 decay_mult: 0 } + param { lr_mult: decay_mult: 1 } param { lr_mult: decay_mult: 0 } convolution_param { num_output: 512 kernel_size: 3 pad: 2 dilation: 2 bottom_is_im2col: true @@ -1119,7 +1123,7 @@ layer { type: "MatWrite" bottom: "fc8_interp" mat_write_param { - prefix: "segaware/features/segaware-all-largeFOV/voc_test/fc8/" + prefix: "segaware/features/vgg/voc_test/fc8/" source: "segaware/list/voc_test_id.txt" strip: 0 period: 1 diff --git a/scripts/segaware/config/vgg/test_voc_val.prototxt b/scripts/segaware/config/vgg/test_voc_val.prototxt index 0dfa8c2..50dce65 100644 --- a/scripts/segaware/config/vgg/test_voc_val.prototxt +++ b/scripts/segaware/config/vgg/test_voc_val.prototxt @@ -16,28 +16,33 @@ # 161; 81; 41; 21; 11 # -name: "segaware-all-largeFOV" +name: "vgg" layer { name: "data" - type: "TwoImageData" + type: "ImageSegData" top: "data" top: "label" + top: "data_dim" + include { + phase: TEST + } image_data_param { - root_folder: "/opt/home/aharley/datasets/VOC2012" - source: "segaware/list/voc_val.txt" + root_folder: "/mnt/data1/kidd/segaware/dataset/VOC2012/VOCtrainval" + source: "/mnt/data1/kidd/segaware/segaware/list/voc_val.txt" batch_size: 1 shuffle: false + label_type: NONE } - two_image_data_param { - first_is_color: true - second_is_color: false - } +#two_image_data_param { +# first_is_color: true +# second_is_color: false +# } transform_param { mean_value: 104.008 mean_value: 116.669 mean_value: 122.675 - crop_size: 513 + crop_size: 321 mirror: false } } @@ -216,8 +221,8 @@ layer { name: "emb22_resized" top: "emb22_resized" interp_param { - height: 513 - width: 513 + height: 321 + width: 321 } } layer { @@ -226,8 +231,8 @@ layer { name: "emb33_resized" top: "emb33_resized" interp_param { - height: 513 - width: 513 + height: 321 + width: 321 } } # Concat, and weight @@ -1123,14 +1128,15 @@ layer { type: "MatWrite" bottom: "fc8_interp" mat_write_param { - prefix: "segaware/features/segaware-all-largeFOV/voc_val/fc8/" - source: "segaware/list/voc_val_id.txt" + prefix: "/mnt/data1/kidd/segaware/segaware/features/vgg/voc_val/fc8/" + source: "/mnt/data1/kidd/segaware/segaware/list/voc_val_id.txt" strip: 0 period: 1 } } layer { bottom: "label" + bottom: "data_dim" name: "silence" type: "Silence" } diff --git a/scripts/segaware/config/vgg/train.prototxt b/scripts/segaware/config/vgg/train.prototxt index 85b616e..c3fc61d 100644 --- a/scripts/segaware/config/vgg/train.prototxt +++ b/scripts/segaware/config/vgg/train.prototxt @@ -20,19 +20,24 @@ name: "${NET_ID}" layer { name: "data" - type: "TwoImageData" + type: "ImageSegData" top: "data" top: "label" + top: "data_dim" + include { + phase: TRAIN + } image_data_param { root_folder: "${DATA_ROOT}" source: "${EXP}/list/${TRAIN_SET}.txt" batch_size: ${BATCH_SIZE} shuffle: true + label_type: PIXEL } - two_image_data_param { - first_is_color: true - second_is_color: false - } +#two_image_data_param { +# first_is_color: true +# second_is_color: false +# } transform_param { mean_value: 104.008 mean_value: 116.669 @@ -1127,3 +1132,8 @@ layer { ignore_label: 255 } } +layer { + name: "silence" + type: "Silence" + bottom: "data_dim" +} diff --git a/scripts/segaware/config/vgg/train_voc_train_aug.prototxt b/scripts/segaware/config/vgg/train_voc_train_aug.prototxt index ca2923a..b2400e5 100644 --- a/scripts/segaware/config/vgg/train_voc_train_aug.prototxt +++ b/scripts/segaware/config/vgg/train_voc_train_aug.prototxt @@ -16,23 +16,28 @@ # 161; 81; 41; 21; 11 # -name: "segaware-all-largeFOV" +name: "vgg" layer { name: "data" - type: "TwoImageData" + type: "ImageSegData" top: "data" top: "label" + top: "data_dim" + include { + phase: TRAIN + } image_data_param { - root_folder: "/opt/home/aharley/datasets/VOC2012" - source: "segaware/list/voc_train_aug.txt" + root_folder: "/mnt/data1/kidd/segaware/dataset/VOC2012/VOCtrainval" + source: "/mnt/data1/kidd/segaware/segaware/list/voc_train_aug.txt" batch_size: 1 shuffle: true + label_type: PIXEL } - two_image_data_param { - first_is_color: true - second_is_color: false - } +#two_image_data_param { +# first_is_color: true +# second_is_color: false +# } transform_param { mean_value: 104.008 mean_value: 116.669 @@ -1127,3 +1132,8 @@ layer { ignore_label: 255 } } +layer { + name: "silence" + type: "Silence" + bottom: "data_dim" +}