CenterNet-Lite
CenterNet-Lite copied to clipboard
yjh0410
mAP of VOC only 0.45?
0.45 mAP?? Incredible！ I got about 74 mAP on VOC2007 test…… And on COCO，it got 25AP and 44.9 AP50……
0.45 mAP?? Incredible！ I got about 74 mAP on VOC2007 test…… And on COCO，it got 25AP and 44.9 AP50……
can you update the scrip of train eval test?
This is my VOC mAP：
VOC07 metric? Yes AP for aeroplane = 0.7701 AP for bicycle = 0.8241 AP for bird = 0.7565 AP for boat = 0.6742 AP for bottle = 0.5818 AP for bus = 0.8644 AP for car = 0.8295 AP for cat = 0.8173 AP for chair = 0.5913 AP for cow = 0.8640 AP for diningtable = 0.6615 AP for dog = 0.8093 AP for horse = 0.8580 AP for motorbike = 0.7935 AP for person = 0.7655 AP for pottedplant = 0.5205 AP for sheep = 0.7897 AP for sofa = 0.6928 AP for train = 0.8282 AP for tvmonitor = 0.7389 Mean AP = 0.7516
You could download my model trained on VOC from BaiduYunDisk： Link：https://pan.baidu.com/s/170OYftGRVW-j5qAKYyHSQQ Password：jz4q
You could download my model trained on VOC from BaiduYunDisk： Link：https://pan.baidu.com/s/170OYftGRVW-j5qAKYyHSQQ Password：jz4q
thanks
my training scrip: python train_voc.py --num_workers 4 --cuda my result: VOC07 metric? Yes AP for aeroplane = 0.5167 AP for bicycle = 0.5855 AP for bird = 0.4687 AP for boat = 0.3199 AP for bottle = 0.2411 AP for bus = 0.4706 AP for car = 0.5975 AP for cat = 0.6274 AP for chair = 0.2386 AP for cow = 0.5675 AP for diningtable = 0.2849 AP for dog = 0.5897 AP for horse = 0.5749 AP for motorbike = 0.5078 AP for person = 0.5153 AP for pottedplant = 0.1922 AP for sheep = 0.5516 AP for sofa = 0.3286 AP for train = 0.4759 AP for tvmonitor = 0.4869 Mean AP = 0.4571
Maybe there is something wrong. But I'm sure that my train, test, eval files are latest. Does the trained model I provide you work? Do you get the same mAP to mine?
Maybe there is something wrong. But I'm sure that my train, test, eval files are latest. Does the trained model I provide you work? Do you get the same mAP to mine?
yes, result: VOC07 metric? Yes AP for aeroplane = 0.7668 AP for bicycle = 0.8239 AP for bird = 0.7580 AP for boat = 0.6738 AP for bottle = 0.5823 AP for bus = 0.8644 AP for car = 0.8283 AP for cat = 0.8186 AP for chair = 0.5939 AP for cow = 0.8640 AP for diningtable = 0.6617 AP for dog = 0.8093 AP for horse = 0.8581 AP for motorbike = 0.7822 AP for person = 0.7669 AP for pottedplant = 0.5250 AP for sheep = 0.7937 AP for sofa = 0.6928 AP for train = 0.8138 AP for tvmonitor = 0.7389 Mean AP = 0.7508
Emmm... Interesting.... Maybe it is because of something random. I plan to train again to check whether I can get 70+ mAP.
Emmm... Interesting.... Maybe it is because of something random. I plan to train again to check whether I can get 70+ mAP.
thanks for your best work, i'm looking for the simple implementation of CenterNet all this time
You're welcome ! My goal is to provide a simple and concise demo of CenterNet. Wish your best !!(●ˇ∀ˇ●)
You're welcome ! My goal is to provide a simple and concise demo of CenterNet. Wish your best !!(●ˇ∀ˇ●)
git clone the latest code, the result is amazing, why? the result: VOC07 metric? Yes AP for aeroplane = 0.7523 AP for bicycle = 0.8118 AP for bird = 0.7611 AP for boat = 0.6420 AP for bottle = 0.5577 AP for bus = 0.8427 AP for car = 0.8335 AP for cat = 0.8483 AP for chair = 0.6042 AP for cow = 0.8525 AP for diningtable = 0.6469 AP for dog = 0.8106 AP for horse = 0.8512 AP for motorbike = 0.8296 AP for person = 0.7603 AP for pottedplant = 0.4943 AP for sheep = 0.7851 AP for sofa = 0.7013 AP for train = 0.8179 AP for tvmonitor = 0.7540 Mean AP = 0.7479
Looks great！ I pushed the latest files yesterday. Maybe there was something wrong in the previous version train file which was ignored.
---Original--- From: "hyaihjq"<[email protected]> Date: Sat, Jul 11, 2020 08:48 AM To: "yjh0410/CenterNet-Lite"<[email protected]>; Cc: "Comment"<[email protected]>;"Jianhua Yang"<[email protected]>; Subject: Re: [yjh0410/CenterNet-Lite] mAP of VOC only 0.45? (#1)
You're welcome ! My goal is to provide a simple and concise demo of CenterNet. Wish your best !!(●ˇ∀ˇ●)
git clone the latest code, the result is amazing, why? the result: VOC07 metric? Yes AP for aeroplane = 0.7523 AP for bicycle = 0.8118 AP for bird = 0.7611 AP for boat = 0.6420 AP for bottle = 0.5577 AP for bus = 0.8427 AP for car = 0.8335 AP for cat = 0.8483 AP for chair = 0.6042 AP for cow = 0.8525 AP for diningtable = 0.6469 AP for dog = 0.8106 AP for horse = 0.8512 AP for motorbike = 0.8296 AP for person = 0.7603 AP for pottedplant = 0.4943 AP for sheep = 0.7851 AP for sofa = 0.7013 AP for train = 0.8179 AP for tvmonitor = 0.7540 Mean AP = 0.7479
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.