iou | Rust interface | Wrapper library

So I'm using the Darknet Framework with YoloV4. This is my yolo_image.py code:

I'm having trouble understanding control flow with asynchronous programming in JS. I come from classic OOP background. eg. C++. Your program starts in the "main" -- top level -- function and it calls other functions to do stuff, but everything always comes back to that main function and it retains overall control. And each sub-function retains control of what they're doing even when they call sub functions. Ultimately the program ends when that main function ends. (That said, that's about as much as I remember of my C++ days so answers with C++ analogies might not be helpful lol).

This makes control flow relatively easy. But I get how that's not designed to handle event driven programming as needed on something like a web server. While Javascript (let's talk node for now, not browser) handles event-driven web servers with callbacks and promises, with relative ease... apparently.

I think I've finally got my head around the idea that with event-driven programming the entry point of the app might do little more than set up a bunch of listeners and then get out of the way (effectively end itself). The listeners pick up all the action and respond.

But sometimes stuff still has to be synchronous, and this is where I keep getting unstuck.

With callbacks, promises, or async/await, we can effectively build synchronous chains of events. eg with Promises:

I am using the Image segmentation guide by fchollet to perform semantic segmentation. I have attempted modifying the guide to suit my dataset by labelling the 8-bit img mask values into 1 and 2 like in the Oxford Pets dataset which will be subtracted to 0 and 1 in class Generator(keras.utils.Sequence).The input image is an RGB-image.

I am not sure why but my dice coefficient isn't increasing at all. I have tried to reduce the learning rate as well as changing the optimizer to SGD/RMSProp, normalizing the data, taking the imbalanced labels into account but the result is very strange. The accuracy/IoU of the model is decreasing as the no. of epochs increases.

If it helps, I previously asked a question about the metrics that I should be using for an imbalanced dataset here. The visualization of the predictions are okay but the metric is not.

What I can do next to debug this problem? Is there anything wrong with my code? Will appreciate any advice.

Here are the results

I am using the Image segmentation guide by fchollet to perform semantic segmentation. I have attempted modifying the guide to suit my dataset by labelling the 8-bit img mask values into 1 and 2 like in the Oxford Pets dataset. (which will be subtracted to 0 and 1 in class OxfordPets(keras.utils.Sequence):)

Question is how do I get the IoU metric of a single class (e.g 1)?

I have tried different metrics suggested by Stack Overflow but most of suggest using MeanIoU which I tried but I have gotten nan loss as a result. Here is an example of a mask after using autocontrast. PIL.ImageOps.autocontrast(load_img(val_target_img_paths[i]))

The model seems to train well but the accuracy was decreasing over time.

Also, can someone help explain how the metric score can be calculated from y_true and y_pred? I don't quite fully understand when the label value is used in the IoU metric calculation.

When I train yolo v4, I get a lot of outputs that I later want to use to plot learning rate changes.

With max_batches = 1543 darknet prints 74064 outputs (with loss etc, I don't count outputs with mAP calculation, network definition etc). With batch = 32, subdivisions = 16, I have only 49376 itarations (defined as single images passed through network)!

Something is definitely off, does anybody know how often this binary is supposed to print values?

I have the fields, group_id, category, account, sku and revenue. I want to do an aggregation where the revenue is grouped by the group_id, category, sku and account separately and in sku there will be two divisions, one where there is a comma and one where there isn't. How do I do this?

For example, if my data is

I have a DataFrame which has Date in it, where the dtype of Date is 'OBJECT'. Now, I need the column names of those which don't have the date in them. I have written code for it but at the end the column names are not getting appended into the list. I really don't know why it is not appending. Can anyone help me with this?

I'm looking for the fastest way to compute the intersection over union (Jaccard Index) of two binary masks (2d arrays of 1s and 0s) in numpy of the exact same shape. My code for computing this is:

I trained multiple models with different configuration for a custom hyperparameter search. I use pytorch_lightning and its logging (TensorboardLogger). When running my training script after Task.init() ClearML auto-creates a Task and connects the logger output to the server.

I log for each straining stage train, val and test the following scalars at each epoch: loss, acc and iou

When I have multiple configuration, e.g. networkA and networkB the first training log its values to loss, acc and iou, but the second to networkB:loss, networkB:acc and networkB:iou. This makes values umcomparable.

My training loop with Task initalization looks like this: