incremental_detectors
incremental_detectors copied to clipboard
kshmelkov
Code release for paper "Incremental Learning of Object Detectors without Catastrophic Forgetting"
I copy this post from [my blog](https://haoyu.love/blog608.html). ----- What is the most distressing thing in life? Not money, not paper, but the author released their fantastic code while you do...
How to get the log/*.log ,please help me.
The mAP on VOC2007 is as follows (19 classes and 19+1 classes): | Category | mAP (40k, 1000000 img) | |---------------+--------------------------| | aeroplane | 3.9 | | bicycle | 9.1...
@kshmelkov Sir, I have trained the network for first 3 classes (GPU 1060 6GB) successfully, but when I am doing the incremental learning with 2 new classes (on GPU 1070...
Hi, Dear Author @kshmelkov Kindly explain the logits than are used as in input to the distillation loss in 2nd training. Please only point out the code that is taking...
Let's see https://github.com/kshmelkov/incremental_detectors/blob/master/network.py#L59-L87 Say `self.num_classes = 15`, and `self.subnet.num_classes = 10`, so we use `logits = self.logits_for_distillation[:, :cached_classes+1]` for distillation. But wait, `[:, :cached_classes+1]`? It is `0-6` now, rather than...
Full error Below usage: frcnn.py [-h] --run_name RUN_NAME [--ckpt CKPT] [--dataset {voc07,voc12,coco}] [--proposals {mcg,edgeboxes,}] --num_classes NUM_CLASSES [--extend EXTEND] [--num_layers NUM_LAYERS] --action ACTION [--data_format {NHWC,NCHW}] [--sigmoid] [--print_step PRINT_STEP] [--conf_thresh CONF_THRESH] [--nms_thresh...
**Hi Konstantin, It is a nice work. And Can you kindly release your COCO model? I just want to finish my master thesis ：） Thank you very much~**
Dear Konstantin, this is a great job for incremental learning. Thanks very much for releasing your code. I just download your code and tried to run them but I am...
i trained the model and got the model.ckpt files (3 of them) . now i want to run inference on custom images to detect objects . how do i do...
