GUIMINLONG

I have a same problem regarding re-constructing pred

> one_hot_predictions, accuracy, final_loss = sess.run( > [pred, accuracy, cost], > feed_dict={ > "x:0": X_test, > "y:0": one_hot(y_test) > } > ) How you get it working?

> I have a related clarification question. What is the proper way of using the offsets computed during the calibration in the setup_MPUs function? > > I have 8 MPU6050s...

The problem already solved. I realize it by replacing the original "mpu.dmp_read_fifo(0);" with "while (!(mpu.dmp_read_fifo(0))){}" 

> Sorry for the delayed post. The code is designed to capture the packets on a first come basis. or to retrieve the packet as soon as it is available...

inverse fcwt will be very helpful ~~~~l

I also wondering how can I reconstruct the signal

that will be very helpful

> _No description provided._ 实现了吗?

> 自定义数据集中包含不同尺寸的图片，并且也不是8的倍数，训练时会报错：  这个有什么解决办法吗?难道需要先对所有数据手动裁剪在进行训练吗？实际情况中输入的每张图片尺寸也不会都是一样的啊 就是要裁剪的吧，模型的框架确定，允许的输入的维度就定死了的。实际情况应该是大了脚本裁剪，小了脚本扩充的吧。