xggIoU / centernet_tensorflow_wilderface_voc

This is the unofficial implementation of the "CenterNet:Objects as Points".Just a simple try with self-modified shufflenetv2 and yolov3.If you want better results, you need more experiments.

MIT License

111 stars 25 forks source link

和论文中不同的地方 #15

Closed dl8207531 closed 5 years ago

dl8207531 commented 5 years ago

请问下这里没用到论文里提到的高斯核和dcn吗？

xggIoU commented 5 years ago

用到了高斯核，以及自己改的高斯核。没有采用dcn。

dl8207531 commented 5 years ago

output_center_peak=tf.layers.max_pooling2d(output_center,3,1,padding='same') peak_mask=tf.cast(tf.equal(output_center,output_center_peak),tf.float32) thresh_mask=tf.cast(tf.greater(output_center,class_prob_thresh),tf.float32) obj_mask=peak_mask*thresh_mask output_center=output_center*obj_mask 看到了是这个吧谢谢

xggIoU commented 5 years ago

这个是最后的预测热图decode操作。为了得到目标的中心点。

dl8207531 commented 5 years ago

是这个函数吗？ gaussian_radius 没找到哪里调用在生成数据的时候使用到的？

xggIoU commented 5 years ago

heatmap_roi = creat_roiheatmap_circle(center_map_obj, size_map_int)

heatmap_roi = creat_roiheatmap_ellipse(center_map_obj, size_map_int)

dl8207531 commented 5 years ago

了解，谢谢