Kamen Rider Black
Results:
Total score: 3/7 available.
Character setting: 2/2
This lora perfectly recreates a uniform that wraps the whole body.
Style: 0/2
The training set may contain only S.H.F. toy photos, or it may not handle the stylized Tags well, resulting in the generated images being all toy photos. A lot of additional tweaks are required to take advantage of the different models.
Fault: 1/2
The marble floor and wall in the background are suspected to be the background of the training set. There are also a small number of special effect props, which may also come from training materials. A small number of physical errors may be due to the fact that all-black clothes are not easily identified as characters.
Subjective: 0/1
A good character Lora should have more possibilities, and I won't pay attention to the Lora trained by the author with this idea at all. PS, I also collected a lot of toys~
Note:
This test is designed to indicate the availability of the character LoRa. Four images were generated using four seeds and ten models.It is evaluated from three dimensions: character setting, style and failure, and one subjective dimension.Objective dimension score: 0 Not Recommended, 1 Available, 2 Recommended,Subjective dimensions: 0 Not recommended, 1 Recommended.Maximum total score of 7 points.
结果:
总分:3/7 可用。
人物设定:2/2
这个lora完美还原出了一套包裹全身的制服。
风格: 0/2
训练集可能只包含S.H.F.玩具照片,也可能是没有处理好风格化的Tag,导致生成图全都是玩具照片。需要很多额外的调整才能发挥不同模型的优势。
故障:1/2
背景中的大理石地面和墙面,怀疑是训练集背景。还有少量特效道具,也可能来源于训练素材。少量肢体错误可能是因为全黑衣服不易被识别为人物。
主观:0/1
好的人物lora应该有更多的可能性,我完全不会再关注作者用这个思路训练出的lora。PS,我也收集了不少玩具~
注意:
该测试旨在指示字符 LoRa 的可用性.使用四个种子和十个模型生成了40个图像。从人物设定、风格与失败三个维度和一个主观维度进行评估。客观维度得分:0 不推荐,1 可用,2 推荐,主观维度:0 不推荐,1 推荐。最高总分 7 分。