Dian-Yi
Dian-Yi
> i find the reason : > cv2.imwrite(str(test_real_dir)+'/{:05}.png'.format(img_idx),img) > cv2.imwrite(str(test_img)+'/{:05}.png'.format(img_idx),img) > cv2.imwrite(str(test_label)+'/{:05}.png'.format(img_idx),label) good job. 不同系统下，不同版本的包，这个项目里一些路径设置的确实有问题，还得慢慢改。
https://github.com/fastai/course-nlp/issues/3 this version maybe ok. but i cant run notebook3 in win10
1. The cache_data() creates file 'cached_data.npy'. I find it is different from file 'BFM_to_FLAME_corr.npz'. So, How to calculate the BFM_to_FLAME_corr.npz? Thanks for your answer. 2. I have a obj model,...
I did some experiments to find problems. There is a sample heatmap label data from: [def generate_label_map(pts, height, width, sigma, downsample, nopoints, ctype):](url) the sum of each map's value >...
have you solved this problem?
Same issue. Maybe someone know how to fix it?
Author says that "Results show that the methods with 95% on raf-db, 80% on affectnet". but I test his model using affectnet, it accuarry is very low.
我简单研究了下track.py里面的one euro filter算法： ``` def __call__(self, x, x_prev): if x_prev is None: return x t_e = 1 a_d = smoothing_factor(t_e, self.d_cutoff) dx = (x - x_prev) / t_e dx_hat =...
请问你解决了这个问题了么？windows下好像没有人脸检测，facebox全是0
Is the training process like deepfakes/swap or deepfacelab? and only swap the trained face in test process?