wzrf
wzrf
Hi, really nice work! I am trying to figure out how to build the jar after I edit the code in the server. Can you help me with it?
Hello! I want to know how can I apply the trained network to images with other resolutions? I notice it's 1266x370, how can I apply this to middlebury dataset?
关于图拓展的问题
想问一下这个图是不是只能一次生成,没法生成之后持续构建?如果要新增文档的话应该怎么做,是不是communities/keywords这些要重新计算?
在函数 _calculate_triple_relevance_scores 中,search出来的D,I中的D本来应该就是similarity(内积相似度分数)把? 之后再用 similarity_score = 1.0 - (distance ** 2) / 2.0 计算是不是算错了? 我打印出来按照similarity_score 排序之后召回的triples 发现好像不太对,辛苦帮忙排查解答下 这是我测试的文档召回recall,其中蓝色是对代码做了如下修改之后: 可以看到召回率有所提升: