Closed whqwill closed 4 years ago
You should copy the fused.csv to hico_benchmark/data/test-result.csv
(if hico_benchmark/data/test-result.csv
already exists, then overwrite it). The matlab code will parse hico_benchmark/data/test-result.csv
for evaluation. The 600-d array indicates the probability that this image contains certain HOI class.
感谢你,但是我们公司禁止使用matlab。代码我看了一下感觉不是很清晰。你能大致说一下代码思路嘛。为什么会有1,0,-1三种标签。
You can regard -1 as negative labels. In fact, -1 comes from the labelings of HICO dataset and it will take very little effects for the final results.
PS: For commercial usage, please email us for an agreement.
It's done. I use octave.
iamge-level中的evaluate用matlab部分中用的测试文件是哪一个。我测试得到的最后结果是linear-final.csv或者fused.csv,但是我看了matlab中的代码的标注label格式是1,0,-1,(分别代表什么意思这里我也不懂),但是linear-final.csv或者fused.csv的格式是一个600维的数组,看起来是激活值,并不是1,0,-1这样的格式,那么最后的结果我怎么得到1,0,-1呢