Faster RCNN model summary for the object detection
Model: "model_5"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to
==================================================================================================
input_1 (InputLayer)            [(None, 400, 300, 3) 0
__________________________________________________________________________________________________
block1_conv1 (Conv2D)           (None, 400, 300, 64) 1792        input_1[0][0]
__________________________________________________________________________________________________
block1_conv2 (Conv2D)           (None, 400, 300, 64) 36928       block1_conv1[0][0]
__________________________________________________________________________________________________
block1_pool (MaxPooling2D)      (None, 200, 150, 64) 0           block1_conv2[0][0]
__________________________________________________________________________________________________
block2_conv1 (Conv2D)           (None, 200, 150, 128 73856       block1_pool[0][0]
__________________________________________________________________________________________________
block2_conv2 (Conv2D)           (None, 200, 150, 128 147584      block2_conv1[0][0]
__________________________________________________________________________________________________
block2_pool (MaxPooling2D)      (None, 100, 75, 128) 0           block2_conv2[0][0]
__________________________________________________________________________________________________
block3_conv1 (Conv2D)           (None, 100, 75, 256) 295168      block2_pool[0][0]
__________________________________________________________________________________________________
block3_conv2 (Conv2D)           (None, 100, 75, 256) 590080      block3_conv1[0][0]
__________________________________________________________________________________________________
block3_conv3 (Conv2D)           (None, 100, 75, 256) 590080      block3_conv2[0][0]
__________________________________________________________________________________________________
block3_pool (MaxPooling2D)      (None, 50, 37, 256)  0           block3_conv3[0][0]
__________________________________________________________________________________________________
block4_conv1 (Conv2D)           (None, 50, 37, 512)  1180160     block3_pool[0][0]
__________________________________________________________________________________________________
block4_conv2 (Conv2D)           (None, 50, 37, 512)  2359808     block4_conv1[0][0]
__________________________________________________________________________________________________
block4_conv3 (Conv2D)           (None, 50, 37, 512)  2359808     block4_conv2[0][0]
__________________________________________________________________________________________________
block4_pool (MaxPooling2D)      (None, 25, 18, 512)  0           block4_conv3[0][0]
__________________________________________________________________________________________________
block5_conv1 (Conv2D)           (None, 25, 18, 512)  2359808     block4_pool[0][0]
__________________________________________________________________________________________________
block5_conv2 (Conv2D)           (None, 25, 18, 512)  2359808     block5_conv1[0][0]
__________________________________________________________________________________________________
block5_conv3 (Conv2D)           (None, 25, 18, 512)  2359808     block5_conv2[0][0]
__________________________________________________________________________________________________
input_2 (InputLayer)            [(None, None, 4)]    0
__________________________________________________________________________________________________
roi_pooling_conv (RoiPoolingCon (1, 4, 7, 7, 512)    0           block5_conv3[0][0]
                                                                 input_2[0][0]
__________________________________________________________________________________________________
time_distributed (TimeDistribut (1, 4, 25088)        0           roi_pooling_conv[0][0]
__________________________________________________________________________________________________
time_distributed_1 (TimeDistrib (1, 4, 4096)         102764544   time_distributed[0][0]
__________________________________________________________________________________________________
time_distributed_2 (TimeDistrib (1, 4, 4096)         0           time_distributed_1[0][0]
__________________________________________________________________________________________________
time_distributed_3 (TimeDistrib (1, 4, 4096)         16781312    time_distributed_2[0][0]
__________________________________________________________________________________________________
rpn_conv1 (Conv2D)              (None, 25, 18, 512)  2359808     block5_conv3[0][0]
__________________________________________________________________________________________________
time_distributed_4 (TimeDistrib (1, 4, 4096)         0           time_distributed_3[0][0]
__________________________________________________________________________________________________
rpn_out_class (Conv2D)          (None, 25, 18, 9)    4617        rpn_conv1[0][0]
__________________________________________________________________________________________________
rpn_out_regress (Conv2D)        (None, 25, 18, 36)   18468       rpn_conv1[0][0]
__________________________________________________________________________________________________
dense_class_38 (TimeDistributed (1, 4, 38)           155686      time_distributed_4[0][0]
__________________________________________________________________________________________________
dense_regress_38 (TimeDistribut (1, 4, 148)          606356      time_distributed_4[0][0]
==================================================================================================
Total params: 137,405,479
Trainable params: 137,405,479
Non-trainable params: 0
__________________________________________________________________________________________________
베이스라인 에러
