import matplotlib.pyplot as plt; 
import cv2
from PIL import Image
import os
import statistics
import multiprocessing as mp
import torch.nn.functional as F
from pathlib import Path
import torch
import random
import imageio
import numpy as np
from glob import glob
from collections import defaultdict
from pdb import set_trace as pdb
from itertools import combinations
from random import choice
import matplotlib.pyplot as plt
import imageio.v3 as iio

from torchvision import transforms

import sys

from glob import glob
import os
import gzip
import json
import numpy as np

from einops import rearrange, repeat

sys.path.append("/home/cameronsmith/repos/multivid_point_track_sfm/")
from data import make_sample

ch_sec = lambda x: rearrange(x,"... c x y -> ... (x y) c")
hom = lambda x, i=-1: torch.cat((x, torch.ones_like(x.unbind(i)[0].unsqueeze(i))), i)

import argparse
parser = argparse.ArgumentParser()
parser.add_argument('-i','--input_dir',  type=str,default="/data/cameron/monocular_ests/re10k/*",required=False,help="dir to pkg")
parser.add_argument('--imgdir',  type=str,default="/data/cameron/monocular_ests/hydrants_redo/*",required=False,help="imgdir to use for focal estr if not already run")
args = parser.parse_args()

from tqdm import tqdm

paths=glob(args.input_dir)
for path in tqdm(paths):
    if os.path.exists(path+"/lowrespkg_flow.pt") and 0:
        print("already done for ",path)
        continue

    if not os.path.exists(path+"/bwd_flow.pt"):
        print("no flow, skipping")
        continue
    imgs = torch.load(path+"/imgs.pt")

    bwd_flow = torch.load(path+"/bwd_flow.pt")
    rig_flow_masks = torch.load(path+"/rig_flow_masks.pt")[:,:1]
    if not os.path.exists(path+"/intrinsics.pt"):
        max_i=40
        print("no intrinsics data, doing")
        image_paths = sorted(Path(args.imgdir).iterdir())[:max_i]
        from geocalib import GeoCalib
        device = "cuda" if torch.cuda.is_available() else "cpu"
        model = GeoCalib().to(device)
        image = model.load_image(image_paths[0]).to(device)
        result = model.calibrate(image, camera_model="pinhole")
        f = result["camera"].f[0,0]/np.array(Image.open(image_paths[0])).shape[1]
        torch.save(f.cpu(),path+"/intrinsics.pt")
    else: f = torch.load(path+"/intrinsics.pt")

    print("Done loading data")
    idx=0

    n_trgt,num_skip,sf=10,3,1#.6

    context = []
    trgt = []
    post_input = []

    frames = imgs
    #f=depths[1]
    #depth_frames = depths[0]

    if frames.max()>2: frames=frames/255

    intrinsics = repeat(torch.eye(3), "i j -> b i j", b=len(imgs)).clone()
    intrinsics[:, :2, 2] = 0.5
    intrinsics[:, 0, 0] = f 
    intrinsics[:, 1, 1] = f * imgs.size(-1) / imgs.size(-2)

    org_ratio=frames[0].size(-2)/frames[0].size(-1)
    h,s=3,1
    hi_res=[640, 1024]

    #s=4
    gs=1

    #rig_flow_masks=torch.ones_like(rig_flow_masks[:,:])
    sample = {
            "intrinsics":intrinsics[:n_trgt*num_skip:num_skip],"rgb":frames[:n_trgt*num_skip:num_skip]* 2-1,
            "org_ratio":org_ratio,
            "rig_flow_masks":rig_flow_masks[:n_trgt*num_skip:num_skip][:-1], 
            "bwd_flow":bwd_flow[:n_trgt*num_skip:num_skip][:-1], 
            }
    switch=[1,-1][0]
    out_dict= make_sample(sample, 1/org_ratio,hires_factor=h,budget=192*640/(8//s),
            low_res=[int(128*sf),int(224*sf)][::switch],#[::[-1,1][frames.size(-1)>frames.size(-2)]],
            hi_res=hi_res[::-1]#[::[-1,1][frames.size(-1)>frames.size(-2)]])
            )
    #if out_dict[0]["rgb"].size(0)!=10: print("skipping x")
    #else: 
    torch.save(out_dict,path+"/lowrespkg_flow.pt")